Network Working Group P. Riikonen
Internet-Draft
draft-riikonen-silc-spec-06.txt 26 November 2002
Expires: 26 April 2003
Secure Internet Live Conferencing (SILC),
Protocol Specification
<draft-riikonen-silc-spec-06.txt>
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC 2026. Internet-Drafts are
working documents of the Internet Engineering Task Force (IETF), its
areas, and its working groups. Note that other groups may also
distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html
The distribution of this memo is unlimited.
Abstract
This memo describes a Secure Internet Live Conferencing (SILC)
protocol which provides secure conferencing services over insecure
network channel. SILC is IRC [IRC] like protocol, however, it is
not equivalent to IRC and does not support IRC. Strong cryptographic
methods are used to protect SILC packets inside the SILC network.
Three other Internet Drafts relates very closely to this memo;
SILC Packet Protocol [SILC2], SILC Key Exchange and Authentication
Protocols [SILC3] and SILC Commands [SILC4].
Riikonen [Page 1]
Internet Draft 26 November 2002
Table of Contents
1 Introduction .................................................. 3
1.1 Requirements Terminology .................................. 4
2 SILC Concepts ................................................. 4
2.1 SILC Network Topology ..................................... 4
2.2 Communication Inside a Cell ............................... 6
2.3 Communication in the Network .............................. 6
2.4 Channel Communication ..................................... 7
2.5 Router Connections ........................................ 8
3 SILC Specification ............................................ 8
3.1 Client .................................................... 9
3.1.1 Client ID ........................................... 9
3.2 Server .................................................... 10
3.2.1 Server's Local ID List .............................. 10
3.2.2 Server ID ........................................... 11
3.2.3 SILC Server Ports ................................... 12
3.3 Router .................................................... 12
3.3.1 Router's Local ID List .............................. 13
3.3.2 Router's Global ID List ............................. 13
3.3.3 Router's Server ID .................................. 14
3.4 Channels .................................................. 14
3.4.1 Channel ID .......................................... 15
3.5 Operators ................................................. 16
3.6 SILC Commands ............................................. 16
3.7 SILC Packets .............................................. 17
3.8 Packet Encryption ......................................... 17
3.8.1 Determination of the Source and the Destination ..... 18
3.8.2 Client To Client .................................... 18
3.8.3 Client To Channel ................................... 20
3.8.4 Server To Server .................................... 20
3.9 Key Exchange And Authentication ........................... 20
3.9.1 Authentication Payload .............................. 21
3.10 Algorithms ............................................... 23
3.10.1 Ciphers ............................................ 23
3.10.1.1 CBC Mode .................................. 24
3.10.1.2 CTR Mode .................................. 24
3.10.1.3 Randomized CBC Mode ....................... 25
3.10.2 Public Key Algorithms .............................. 26
3.10.3 Hash Functions ..................................... 26
3.10.4 MAC Algorithms ..................................... 27
3.10.5 Compression Algorithms ............................. 27
3.11 SILC Public Key .......................................... 28
3.12 SILC Version Detection ................................... 30
3.13 Backup Routers ........................................... 31
3.13.1 Switching to Backup Router ......................... 32
3.13.2 Resuming Primary Router ............................ 33
3.13.3 Discussion on Backup Router Scheme ................. 36
Riikonen [Page 2]
Internet Draft 26 November 2002
4 SILC Procedures ............................................... 36
4.1 Creating Client Connection ................................ 36
4.2 Creating Server Connection ................................ 38
4.2.1 Announcing Clients, Channels and Servers ............ 38
4.3 Joining to a Channel ...................................... 39
4.4 Channel Key Generation .................................... 41
4.5 Private Message Sending and Reception ..................... 41
4.6 Private Message Key Generation ............................ 42
4.7 Channel Message Sending and Reception ..................... 43
4.8 Session Key Regeneration .................................. 43
4.9 Command Sending and Reception ............................. 44
4.10 Closing Connection ....................................... 45
4.11 Detaching and Resuming a Session ......................... 45
5 Security Considerations ....................................... 47
6 References .................................................... 48
7 Author's Address .............................................. 49
List of Figures
Figure 1: SILC Network Topology
Figure 2: Communication Inside cell
Figure 3: Communication Between Cells
Figure 4: Router Connections
Figure 5: SILC Public Key
Figure 6: Counter Block
1. Introduction
This document describes a Secure Internet Live Conferencing (SILC)
protocol which provides secure conferencing services over insecure
network channel. SILC is IRC [IRC] like protocol, however, it is
not equivalent to IRC and does not support IRC. Some of the SILC's
features are not found in IRC but in traditional Instant Message (IM)
protocols. SILC combines features from both of these chat protocol
styles, and SILC can be implemented as either IRC-like system or
IM-like system.
Strong cryptographic methods are used to protect SILC packets inside
the SILC network. Three other Internet Drafts relates very closely
to this memo; SILC Packet Protocol [SILC2], SILC Key Exchange and
Authentication Protocols [SILC3] and SILC Commands [SILC4].
The protocol uses extensively packets as conferencing protocol
requires message and command sending. The SILC Packet Protocol is
described in [SILC2] and should be read to fully comprehend this
document and protocol. [SILC2] also describes the packet encryption
Riikonen [Page 3]
Internet Draft 26 November 2002
and decryption in detail. The SILC Packet Protocol provides secured
and authenticated packets, and the protocol is designed to be compact.
This makes SILC also suitable in environment of low bandwidth
requirements such as mobile networks. All packet payloads in SILC
can be also compressed.
The security of SILC protocol, and for any security protocol for that
matter, is based on strong and secure key exchange protocol. The SILC
Key Exchange protocol is described in [SILC3] along with connection
authentication protocol and should be read to fully comprehend this
document and protocol.
The SILC protocol has been developed to work on TCP/IP network
protocol, although it could be made to work on other network protocols
with only minor changes. However, it is recommended that TCP/IP
protocol is used under SILC protocol. Typical implementation would
be made in client-server model.
1.1 Requirements Terminology
The keywords MUST, MUST NOT, REQUIRED, SHOULD, SHOULD NOT, RECOMMENDED,
MAY, and OPTIONAL, when they appear in this document, are to be
interpreted as described in [RFC2119].
2. SILC Concepts
This section describes various SILC protocol concepts that forms the
actual protocol, and in the end, the actual SILC network. The mission
of the protocol is to deliver messages from clients to other clients
through routers and servers in secure manner. The messages may also
be delivered from one client to many clients forming a group, also
known as a channel.
This section does not focus to security issues. Instead, basic network
concepts are introduced to make the topology of the SILC network
clear.
2.1 SILC Network Topology
SILC network forms a ring as opposed to tree style network topology that
conferencing protocols usually have. The network has a cells which are
constructed from router and zero or more servers. The servers are
connected to the router in a star like network topology. Routers in the
network are connected to each other forming a ring. The rationale for
this is to have servers that can perform specific kind of tasks what
Riikonen [Page 4]
Internet Draft 26 November 2002
other servers cannot perform. This leads to two kinds of servers; normal
SILC servers and SILC routers.
A difference between normal server and router server is that routers
knows everything about everything in the network. They also do the
actual routing of the messages to the correct receiver. Normal servers
knows only about local information and nothing about global information.
This makes the network faster as there are less servers that needs to
keep global information up to date at all time.
This, on the other hand, leads to kind of a cellular like network, where
routers are in the center of the cell and servers are connected to the
router.
The following diagram represents SILC network topology.
---- ---- ---- ---- ---- ----
| S8 | S5 | S4 | | S7 | S5 | S6 |
----- ---- ----- ----- ---- -----
| S7 | S/R1 | S2 | --- | S8 | S/R2 | S4 |
---- ------ ---- ---- ------ ----
| S6 | S3 | S1 | | S1 | S3 | S2 | ---- ----
---- ---- ---- ---- ---- ---- | S3 | S1 |
Cell 1. \ Cell 2. | \____ ----- -----
| | | S4 | S/R4 |
---- ---- ---- ---- ---- ---- ---- ------
| S7 | S4 | S2 | | S1 | S3 | S2 | | S2 | S5 |
----- ---- ----- ----- ---- ----- ---- ----
| S6 | S/R3 | S1 | --- | S4 | S/R5 | S5 | ____/ Cell 4.
---- ------ ---- ---- ------ ----
| S8 | S5 | S3 | | S6 | S7 | S8 | ... etc ...
---- ---- ---- ---- ---- ----
Cell 3. Cell 5.
Figure 1: SILC Network Topology
A cell is formed when a server or servers connect to one router. In
SILC network normal server cannot directly connect to other normal
server. Normal server may only connect to SILC router which then
routes the messages to the other servers in the cell. Router servers
on the other hand may connect to other routers to form the actual SILC
network, as seen in above figure. However, router is also normal SILC
server; clients may connect to it the same way as to normal SILC
server. Normal server also cannot have active connections to more
than one router. Normal server cannot be connected to two different
cells. Router servers, on the other hand, may have as many router to
router connections as needed.
Riikonen [Page 5]
Internet Draft 26 November 2002
There are many issues in this network topology that needs to be careful
about. Issues like the size of the cells, the number of the routers in
the SILC network and the capacity requirements of the routers. These
issues should be discussed in the Internet Community and additional
documents on the issue may be written.
2.2 Communication Inside a Cell
It is always guaranteed that inside a cell message is delivered to the
recipient with at most two server hops. A client which is connected to
server in the cell and is talking on channel to other client connected
to other server in the same cell, will have its messages delivered from
its local server first to the router of the cell, and from the router
to the other server in the cell.
The following diagram represents this scenario:
1 --- S1 S4 --- 5
S/R
2 -- S2 S3
/ |
4 3
Figure 2: Communication Inside cell
Example: Client 1. connected to Server 1. send message to
Client 4. connected to Server 2. travels from Server 1.
first to Router which routes the message to Server 2.
which then sends it to the Client 4. All the other
servers in the cell will not see the routed message.
If the client is connected directly to the router, as router is also normal
SILC server, the messages inside the cell are always delivered only with
one server hop. If clients communicating with each other are connected
to the same server, no router interaction is needed. This is the optimal
situation of message delivery in the SILC network.
2.3 Communication in the Network
If the message is destined to server that does not belong to local cell
the message is routed to the router server to which the destination
server belongs, if the local router is connected to destination router.
Riikonen [Page 6]
Internet Draft 26 November 2002
If there is no direct connection to the destination router, the local
router routes the message to its primary route. The following diagram
represents message sending between cells.
1 --- S1 S4 --- 5 S2 --- 1
S/R - - - - - - - - S/R
2 -- S2 S3 S1
/ | \
4 3 2
Cell 1. Cell 2.
Figure 3: Communication Between Cells
Example: Client 5. connected to Server 4. in Cell 1. sends message
to Client 2. connected to Server 1. in Cell 2. travels
from Server 4. to Router which routes the message to
Router in Cell 2, which then routes the message to
Server 1. All the other servers and routers in the
network will not see the routed message.
The optimal case of message delivery from the client point of view is
when clients are connected directly to the routers and the messages
are delivered from one router to the other.
2.4 Channel Communication
Messages may be sent to group of clients as well. Sending messages to
many clients works the same way as sending messages point to point, from
message delivery point of view. Security issues are another matter
which are not discussed in this section.
Router server handles the message routing to multiple recipients. If
any recipient is not in the same cell as the sender the messages are
routed further.
Server distributes the channel message to its local clients which are
joined to the channel. Router also distributes the message to its
local clients on the channel.
Riikonen [Page 7]
Internet Draft 26 November 2002
2.5 Router Connections
Router connections play very important role in making the SILC like
network topology to work. For example, sending broadcast packets in
SILC network require special connections between routers; routers must
be connected in a specific way.
Every router has their primary route which is a connection to another
router in the network. Unless there is only two routers in the network
must not routers use each other as their primary routes. The router
connections in the network must form a ring.
Example with three routers in the network:
S/R1 - < - < - < - < - < - < - S/R2
\ /
v ^
\ - > - > - S/R3 - > - > - /
Figure 4: Router Connections
Example: Network with three routers. Router 1. uses Router 2. as its
primary router. Router 2. uses Router 3. as its primary router,
and Router 3. uses Router 1. as its primary router. There may
be other direct connections between the routers but they must
not be used as primary routes.
The above example is applicable to any amount of routers in the network
except for two routers. If there are only two routers in the network both
routers must be able to handle situation where they use each other as their
primary routes.
The issue of router connections are very important especially with SILC
broadcast packets. Usually all router wide information in the network is
distributed by SILC broadcast packets. This sort of ring network, with
ability to have other direct routes in the network cause interesting
routing problems. The [SILC2] discusses the routing of packets in this
sort of network in more detail.
3. SILC Specification
This section describes the SILC protocol. However, [SILC2] and
[SILC3] describes other important protocols that are part of this SILC
specification and must be read.
Riikonen [Page 8]
Internet Draft 26 November 2002
3.1 Client
A client is a piece of software connecting to SILC server. SILC client
cannot be SILC server. Purpose of clients is to provide the user
interface of the SILC services for end user. Clients are distinguished
from other clients by unique Client ID. Client ID is a 128 bit ID that
is used in the communication in the SILC network. The client ID is
based on the nickname selected by the user. User uses logical nicknames
in communication which are then mapped to the corresponding Client ID.
Client ID's are low level identifications and should not be seen by the
end user.
Clients provide other information about the end user as well. Information
such as the nickname of the user, username and the host name of the end
user and user's real name. See section 3.2 Server for information of
the requirements of keeping this information.
The nickname selected by the user is not unique in the SILC network.
There can be 2^8 same nicknames for one IP address. As for comparison to
IRC [IRC] where nicknames are unique this is a fundamental difference
between SILC and IRC. This typically causes the server names or client's
host names to be used along with the nicknames on user interface to
identify specific users when sending messages. This feature of SILC
makes IRC style nickname-wars obsolete as no one owns their nickname;
there can always be someone else with the same nickname. The maximum
length of nickname is 128 bytes.
3.1.1 Client ID
Client ID is used to identify users in the SILC network. The Client ID
is unique to the extent that there can be 2^128 different Client ID's,
and ID's based on IPv6 addresses extends this to 2^224 different Client
ID's. Collisions are not expected to happen. The Client ID is defined
as follows.
128 bit Client ID based on IPv4 addresses:
32 bit Server ID IP address (bits 1-32)
8 bit Random number or counter
88 bit Truncated MD5 hash value of the nickname
224 bit Client ID based on IPv6 addresses:
128 bit Server ID IP address (bits 1-128)
8 bit Random number or counter
88 bit Truncated MD5 hash value of the nickname
Riikonen [Page 9]
Internet Draft 26 November 2002
o Server ID IP address - Indicates the server where this
client is coming from. The IP address hence equals the
server IP address where to the client has connected.
o Random number or counter - Random number to further
randomize the Client ID. Another choice is to use
a counter starting from the zero (0). This makes it
possible to have 2^8 same nicknames from the same
server IP address.
o MD5 hash - MD5 hash value of the lowercase nickname is
truncated taking 88 bits from the start of the hash value.
This hash value is used to search the user's Client ID
from the ID lists. Note that the nickname MUST be in
lowercase format.
Collisions could occur when more than 2^8 clients using same nickname
from the same server IP address is connected to the SILC network.
Server MUST be able to handle this situation by refusing to accept
anymore of that nickname.
Another possible collision may happen with the truncated hash value of
the nickname. It could be possible to have same truncated hash value for
two different nicknames. However, this is not expected to happen nor
cause any problems if it would occur. Nicknames are usually logical and
it is unlikely to have two distinct logical nicknames produce same
truncated hash value.
3.2 Server
Servers are the most important parts of the SILC network. They form the
basis of the SILC, providing a point to which clients may connect to.
There are two kinds of servers in SILC; normal servers and router servers.
This section focus on the normal server and router server is described
in the section 3.3 Router.
Normal servers MUST NOT directly connect to other normal server. Normal
servers may only directly connect to router server. If the message sent
by the client is destined outside the local server it is always sent to
the router server for further routing. Server may only have one active
connection to router on same port. Normal server MUST NOT connect to other
cell's router except in situations where its cell's router is unavailable.
3.2.1 Server's Local ID List
Normal server keeps various information about the clients and their end
Riikonen [Page 10]
Internet Draft 26 November 2002
users connected to it. Every normal server MUST keep list of all locally
connected clients, Client ID's, nicknames, usernames and host names and
user's real name. Normal servers only keeps local information and it
does not keep any global information. Hence, normal servers knows only
about their locally connected clients. This makes servers efficient as
they don't have to worry about global clients. Server is also responsible
of creating the Client ID's for their clients.
Normal server also keeps information about locally created channels and
their Channel ID's.
Hence, local list for normal server includes:
server list - Router connection
o Server name
o Server IP address
o Server ID
o Sending key
o Receiving key
o Public key
client list - All clients in server
o Nickname
o Username@host
o Real name
o Client ID
o Sending key
o Receiving key
o Public key
channel list - All channels in server
o Channel name
o Channel ID
o Client ID's on channel
o Client ID modes on channel
o Channel key
3.2.2 Server ID
Servers are distinguished from other servers by unique 64 bit Server ID
(for IPv4) or 160 bit Server ID (for IPv6). The Server ID is used in
the SILC to route messages to correct servers. Server ID's also provide
information for Client ID's, see section 3.1.1 Client ID. Server ID is
defined as follows.
64 bit Server ID based on IPv4 addresses:
Riikonen [Page 11]
Internet Draft 26 November 2002
32 bit IP address of the server
16 bit Port
16 bit Random number
160 bit Server ID based on IPv6 addresses:
128 bit IP address of the server
16 bit Port
16 bit Random number
o IP address of the server - This is the real IP address of
the server.
o Port - This is the port the server is bound to.
o Random number - This is used to further randomize the Server ID.
Collisions are not expected to happen in any conditions. The Server ID
is always created by the server itself and server is responsible of
distributing it to the router.
3.2.3 SILC Server Ports
The following ports has been assigned by IANA for the SILC protocol:
silc 706/tcp SILC
silc 706/udp SILC
If there are needs to create new SILC networks in the future the port
numbers must be officially assigned by the IANA.
Server on network above privileged ports (>1023) SHOULD NOT be trusted
as they could have been set up by untrusted party.
3.3 Router
Router server in SILC network is responsible for keeping the cell together
and routing messages to other servers and to other routers. Router server
is also a normal server thus clients may connect to it as it would be
just normal SILC server.
However, router servers has a lot of important tasks that normal servers
do not have. Router server knows everything about everything in the SILC.
They know all clients currently on SILC, all servers and routers and all
channels in SILC. Routers are the only servers in SILC that care about
Riikonen [Page 12]
Internet Draft 26 November 2002
global information and keeping them up to date at all time. And, this
is what they must do.
3.3.1 Router's Local ID List
Router server as well MUST keep local list of connected clients and
locally created channels. However, this list is extended to include all
the informations of the entire cell, not just the server itself as for
normal servers.
However, on router this list is a lot smaller since routers do not need
to keep information about user's nickname, username and host name and real
name since these are not needed by the router. The router keeps only
information that it needs.
Hence, local list for router includes:
server list - All servers in the cell
o Server name
o Server ID
o Router's Server ID
o Sending key
o Receiving key
client list - All clients in the cell
o Client ID
channel list - All channels in the cell
o Channel ID
o Client ID's on channel
o Client ID modes on channel
o Channel key
Note that locally connected clients and other information include all the
same information as defined in section section 3.2.1 Server's Local ID
List.
3.3.2 Router's Global ID List
Router server MUST also keep global list. Normal servers do not have
global list as they know only about local information. Global list
includes all the clients on SILC, their Client ID's, all created channels
and their Channel ID's and all servers and routers on SILC and their
Server ID's. That is said, global list is for global information and the
list must not include the local information already on the router's local
Riikonen [Page 13]
Internet Draft 26 November 2002
list.
Note that the global list does not include information like nicknames,
usernames and host names or user's real names. Router does not need to
keep these informations as they are not needed by the router. This
information is available from the client's server which maybe queried
when needed.
Hence, global list includes:
server list - All servers in SILC
o Server name
o Server ID
o Router's Server ID
client list - All clients in SILC
o Client ID
channel list - All channels in SILC
o Channel ID
o Client ID's on channel
o Client ID modes on channel
3.3.3 Router's Server ID
Router's Server ID's are equivalent to normal Server ID's. As routers
are normal servers as well same types of ID's applies for routers as well.
Thus, see section 3.2.2 Server ID.
3.4 Channels
A channel is a named group of one or more clients which will all receive
messages addressed to that channel. The channel is created when first
client requests JOIN command to the channel, and the channel ceases to
exist when the last client has left it. When channel exists, any client
can reference it using the name of the channel. If the channel has
a founder mode set and last client leaves the channel the channel does
not cease to exist. The founder mode can be used to make permanent
channels in the network. The founder of the channel can regain the
channel founder privileges on the channel later when he joins the
channel.
Channel names are unique although the real uniqueness comes from 64 bit
Channel ID. However, channel names are still unique and no two global
channels with same name may exist. The channel name is a string of
Riikonen [Page 14]
Internet Draft 26 November 2002
maximum length of 256 bytes. Channel names MUST NOT contain any
whitespaces (` '), any non-printable ASCII characters, commas (`,')
and wildcard characters.
Channels can have operators that can administrate the channel and
operate all of its modes. The following operators on channel exist on
the SILC network.
o Channel founder - When channel is created the joining client becomes
channel founder. Channel founder is channel operator with some more
privileges. Basically, channel founder can fully operate the channel
and all of its modes. The privileges are limited only to the
particular channel. There can be only one channel founder per
channel. Channel founder supersedes channel operator's privileges.
Channel founder privileges cannot be removed by any other operator on
channel. When channel founder leaves the channel there is no channel
founder on the channel. However, it is possible to set a mode for
the channel which allows the original channel founder to regain the
founder privileges even after leaving the channel. Channel founder
also cannot be removed by force from the channel.
o Channel operator - When client joins to channel that has not existed
previously it will become automatically channel operator (and channel
founder discussed above). Channel operator is able administrate the
channel, set some modes on channel, remove a badly behaving client
from the channel and promote other clients to become channel
operator. The privileges are limited only to the particular channel.
Normal channel user may be promoted (opped) to channel operator
gaining channel operator privileges. Channel founder or other
channel operator may also demote (deop) channel operator to normal
channel user.
3.4.1 Channel ID
Channels are distinguished from other channels by unique Channel ID.
The Channel ID is a 64 bit ID (for IPv4) or 160 bit ID (for IPv6), and
collisions are not expected to happen in any conditions. Channel names
are just for logical use of channels. The Channel ID is created by the
server where the channel is created. The Channel ID is defined as
follows.
64 bit Channel ID based on IPv4 addresses:
32 bit Router's Server ID IP address (bits 1-32)
16 bit Router's Server ID port (bits 33-48)
Riikonen [Page 15]
Internet Draft 26 November 2002
16 bit Random number
160 bit Channel ID based on IPv6 addresses:
128 bit Router's Server ID IP address (bits 1-128)
16 bit Router's Server ID port (bits 129-144)
16 bit Random number
o Router's Server ID IP address - Indicates the IP address of
the router of the cell where this channel is created. This is
taken from the router's Server ID. This way SILC router knows
where this channel resides in the SILC network.
o Router's Server ID port - Indicates the port of the channel on
the server. This is taken from the router's Server ID.
o Random number - To further randomize the Channel ID. This makes
sure that there are no collisions. This also means that
in a cell there can be 2^16 channels.
3.5 Operators
Operators are normal users with extra privileges to their server or
router. Usually these people are SILC server and router administrators
that take care of their own server and clients on them. The purpose of
operators is to administrate the SILC server or router. However, even
an operator with highest privileges is not able to enter invite-only
channel, to gain access to the contents of a encrypted and authenticated
packets traveling in the SILC network or to gain channel operator
privileges on public channels without being promoted. They have the
same privileges as everyone else except they are able to administrate
their server or router.
3.6 SILC Commands
Commands are very important part on SILC network especially for client
which uses commands to operate on the SILC network. Commands are used
to set nickname, join to channel, change modes and many other things.
Client usually sends the commands and server replies by sending a reply
packet to the command. Server MAY also send commands usually to serve
the original client's request. Usually server cannot send commands to
clients, however there MAY be commands that allow the server to send
commands to client. By default servers MAY send commands only to other
servers and routers.
Riikonen [Page 16]
Internet Draft 26 November 2002
Note that the command reply is usually sent only after client has sent
the command request but server is allowed to send command reply packet
to client even if client has not requested the command. Client MAY
choose to ignore the command reply.
It is expected that some of the commands may be miss-used by clients
resulting various problems on the server side. Every implementation
SHOULD assure that commands may not be executed more than once, say,
in two (2) seconds. However, to keep response rate up, allowing for
example five (5) commands before limiting is allowed. It is RECOMMENDED
that commands such as SILC_COMMAND_NICK, SILC_COMMAND_JOIN,
SILC_COMMAND_LEAVE and SILC_COMMAND_KILL SHOULD be limited in all cases
as they require heavy operations. This should be sufficient to prevent
the miss-use of commands.
SILC commands are described in [SILC4].
3.7 SILC Packets
Packets are naturally the most important part of the protocol and the
packets are what actually makes the protocol. Packets in SILC network
are always encrypted using, usually the shared secret session key
or some other key, for example, channel key, when encrypting channel
messages. It is not possible to send packet in SILC network without
encryption. The SILC Packet Protocol is a wide protocol and is described
in [SILC2]. This document does not define or describe details of
SILC packets.
3.8 Packet Encryption
All packets passed in SILC network MUST be encrypted. This section
defines how packets must be encrypted in the SILC network. The detailed
description of the actual encryption process of the packets are
described in [SILC2].
Client and its server shares secret symmetric session key which is
established by the SILC Key Exchange Protocol, described in [SILC3].
Every packet sent from client to server, with exception of packets for
channels, are encrypted with this session key.
Channels has a channel key that are shared by every client on the channel.
However, the channel keys are cell specific thus one cell does not know
the channel key of the other cell, even if that key is for same channel.
Channel key is also known by the routers and all servers that has clients
on the channel. However, channels MAY have channel private keys that
are entirely local setting for the client. All clients on the channel
Riikonen [Page 17]
Internet Draft 26 November 2002
MUST know the channel private key before hand to be able to talk on the
channel. In this case, no server or router know the key for channel.
Server shares secret symmetric session key with router which is
established by the SILC Key Exchange Protocol. Every packet passed from
server to router, with exception of packets for channels, are encrypted
with the shared session key. Same way, router server shares secret
symmetric key with its primary route. However, every packet passed
from router to other router, including packets for channels, are
encrypted with the shared session key. Every router connection has
their own session keys.
3.8.1 Determination of the Source and the Destination
The source and the destination of the packet needs to be determined
to be able to route the packets to correct receiver. This information
is available in the SILC Packet Header which is included in all packets
sent in SILC network. The SILC Packet Header is described in [SILC2].
The header MUST be encrypted with the session key who is next receiver
of the packet along the route. The receiver of the packet, for example
a router along the route, is able to determine the sender and the
destination of the packet by decrypting the SILC Packet Header and
checking the ID's attached to the header. The ID's in the header will
tell to where the packet needs to be sent and where it is coming from.
The header in the packet MUST NOT change during the routing of the
packet. The original sender, for example client, assembles the packet
and the packet header and server or router between the sender and the
receiver MUST NOT change the packet header. Note however, that some
packets such as commands may be resent by a server to serve the client's
original command. In this case the command packet sent by the server
includes the server's IDs.
Note that the packet and the packet header may be encrypted with
different keys. For example, packets to channels are encrypted with
the channel key, however, the header is encrypted with the session key
as described above. However, the header and the packet may be encrypted
with same key. This is the case, for example, with command packets.
3.8.2 Client To Client
The process of message delivery and encryption from client to another
client is as follows.
Example: Private message from client to another client on different
Riikonen [Page 18]
Internet Draft 26 November 2002
servers. Clients do not share private message delivery
keys; normal session keys are used.
o Client 1. sends encrypted packet to its server. The packet is
encrypted with the session key shared between client and its
server.
o Server determines the destination of the packet and decrypts
the packet. Server encrypts the packet with session key shared
between the server and its router, and sends the packet to the
router.
o Router determines the destination of the packet and decrypts
the packet. Router encrypts the packet with session key
shared between the router and the destination server, and sends
the packet to the server.
o Server determines the client to which the packet is destined
to and decrypts the packet. Server encrypts the packet with
session key shared between the server and the destination client,
and sends the packet to the client.
o Client 2. decrypts the packet.
Example: Private message from client to another client on different
servers. Clients has established secret shared private
message delivery key with each other and that is used in
the message encryption.
o Client 1. sends encrypted packet to its server. The packet header
is encrypted with the session key shared between the client and
server, and the private message is encrypted with the private
message delivery key shared between clients.
o Server determines the destination of the packet and sends the
packet to the router.
o Router determines the destination of the packet and sends the
packet to the server.
o Server determines the client to which the packet is destined
to and sends the packet to the client.
o Client 2. decrypts the packet with the secret shared key.
If clients share secret key with each other the private message
delivery is much simpler since servers and routers between the
Riikonen [Page 19]
Internet Draft 26 November 2002
clients do not need to decrypt and re-encrypt the packet.
The process for clients on same server is much simpler as there are
no need to send the packet to the router. The process for clients
on different cells is same as above except that the packet is routed
outside the cell. The router of the destination cell routes the
packet to the destination same way as described above.
3.8.3 Client To Channel
Process of message delivery from client on channel to all the clients
on the channel.
Example: Channel of four users; two on same server, other two on
different cells. Client sends message to the channel.
o Client 1. encrypts the packet with channel key and sends the
packet to its server.
o Server determines local clients on the channel and sends the
packet to the Client on the same server. Server then sends
the packet to its router for further routing.
o Router determines local clients on the channel, if found
sends packet to the local clients. Router determines global
clients on the channel and sends the packet to its primary
router or fastest route.
o (Other router(s) do the same thing and sends the packet to
the server(s))
o Server determines local clients on the channel and sends the
packet to the client.
o All clients receiving the packet decrypts it.
3.8.4 Server To Server
Server to server packet delivery and encryption is described in above
examples. Router to router packet delivery is analogous to server to
server. However, some packets, such as channel packets, are processed
differently. These cases are described later in this document and
more in detail in [SILC2].
3.9 Key Exchange And Authentication
Riikonen [Page 20]
Internet Draft 26 November 2002
Key exchange is done always when for example client connects to server
but also when server and router, and router and router connects to each
other. The purpose of key exchange protocol is to provide secure key
material to be used in the communication. The key material is used to
derive various security parameters used to secure SILC packets. The
SILC Key Exchange protocol is described in detail in [SILC3].
Authentication is done after key exchange protocol has been successfully
completed. The purpose of authentication is to authenticate for example
client connecting to the server. However, clients may be accepted
to connect to server without explicit authentication. Servers are
required to use authentication protocol when connecting. The
authentication may be based on passphrase (pre-shared-secret) or public
key based on digital signatures. All passphrases sent in SILC protocol
MUST be UTF-8 [RFC2279] encoded. The connection authentication protocol
is described in detail in [SILC3].
3.9.1 Authentication Payload
Authentication payload is used separately from the SKE and the Connection
Authentication protocol. It can be used during the session to authenticate
with the remote. For example, the client can authenticate itself to the
server to become server operator. In this case, Authentication Payload is
used.
The format of the Authentication Payload is as follows:
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length | Authentication Method |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Public Data Length | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| |
~ Public Data ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Authentication Data Length | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| |
~ Authentication Data ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 5: Authentication Payload
Riikonen [Page 21]
Internet Draft 26 November 2002
o Payload Length (2 bytes) - Length of the entire payload.
o Authentication Method (2 bytes) - The method of the
authentication. The authentication methods are defined
in [SILC2] in the Connection Auth Request Payload. The NONE
authentication method SHOULD NOT be used.
o Public Data Length (2 bytes) - Indicates the length of
the Public Data field.
o Public Data (variable length) - This is defined only if
the authentication method is public key. If it is any other
this field MAY include a random data for padding purposes.
However, in this case the field MUST be ignored by the
receiver.
When the authentication method is public key this includes
128 to 4096 bytes of non-zero random data that is used in
the signature process, described subsequently.
o Authentication Data Length (2 bytes) - Indicates the
length of the Authentication Data field. If zero (0)
value is found in this field the payload MUST be
discarded.
o Authentication Data (variable length) - Authentication
method dependent authentication data.
If the authentication method is password based, the Authentication
Data field includes the plaintext UTF-8 encoded password. It is safe
to send plaintext password since the entire payload is encrypted. In
this case the Public Data Length is set to zero (0), but MAY also include
random data for padding purposes. It is also RECOMMENDED that maximum
amount of padding is applied to SILC packet when using password based
authentication. This way it is not possible to approximate the length
of the password from the encrypted packet.
If the authentication method is public key based (or certificate)
the Authentication Data is computed as follows:
HASH = hash(random bytes | ID | public key (or certificate));
Authentication Data = sign(HASH);
The hash() and the sign() are the hash function and the public key
cryptography function selected in the SKE protocol, unless otherwise
stated in the context where this payload is used. The public key
is SILC style public key unless certificates are used. The ID is the
Riikonen [Page 22]
Internet Draft 26 November 2002
entity's ID (Client or Server ID) which is authenticating itself. The
ID encoding is described in [SILC2]. The random bytes are non-zero
random bytes of length between 128 and 4096 bytes, and will be included
into the Public Data field as is.
The receiver will compute the signature using the random data received
in the payload, the ID associated to the connection and the public key
(or certificate) received in the SKE protocol. After computing the
receiver MUST verify the signature. In case of public key authentication
also this payload is encrypted.
3.10 Algorithms
This section defines all the allowed algorithms that can be used in
the SILC protocol. This includes mandatory cipher, mandatory public
key algorithm and MAC algorithms.
3.10.1 Ciphers
Cipher is the encryption algorithm that is used to protect the data
in the SILC packets. See [SILC2] of the actual encryption process and
definition of how it must be done. SILC has a mandatory algorithm that
must be supported in order to be compliant with this protocol.
The following ciphers are defined in SILC protocol:
aes-256-cbc AES in CBC mode, 256 bit key (REQUIRED)
aes-256-ctr AES in CTR mode, 256 bit key (RECOMMENDED)
aes-256-rcbc AES in randomized CBC mode, 256 bit key (OPTIONAL)
aes-192-<mode> AES in <mode> mode, 192 bit key (OPTIONAL)
aes-128-<mode> AES in <mode> mode, 128 bit key (RECOMMENDED)
twofish-256-<mode> Twofish in <mode> mode, 256 bit key (OPTIONAL)
twofish-192-<mode> Twofish in <mode> mode, 192 bit key (OPTIONAL)
twofish-128-<mode> Twofish in <mode> mode, 128 bit key (OPTIONAL)
cast-256-<mode> CAST-256 in <mode> mode, 256 bit key (OPTIONAL)
cast-192-<mode> CAST-256 in <mode> mode, 192 bit key (OPTIONAL)
cast-128-<mode> CAST-256 in <mode> mode, 128 bit key (OPTIONAL)
serpent-<len>-<mode> Serpent in <mode> mode, <len> bit key (OPTIONAL)
rc6-<len>-<mode> RC6 in <mode> mode, <len> bit key (OPTIONAL)
mars-<len>-<mode> MARS in <mode> mode, <len> bit key (OPTIONAL)
none No encryption (OPTIONAL)
The <mode> is either "cbc", "ctr" or "rcbc". Other encryption modes MAY
be defined as to be used in SILC using the same format. The <len> is
either 256, 192 or 128 bit key length. Also, additional ciphers MAY be
defined to be used in SILC by using the same name format as above.
Riikonen [Page 23]
Internet Draft 26 November 2002
Algorithm "none" does not perform any encryption process at all and
thus is not recommended to be used. It is recommended that no client
or server implementation would accept none algorithm except in special
debugging mode.
3.10.1.1 CBC Mode
The "cbc" encryption mode is CBC mode with inter-packet chaining. This
means that the Initial Vector (IV) for the next encryption block is
the previous ciphertext block. The very first IV MUST be random and is
generated as described in [SILC3].
3.10.1.2 CTR Mode
The "ctr" encryption mode is CTR mode. The CTR mode in SILC is stateful
in encryption and decryption. Both sender and receiver maintain the
counter for the CTR mode and thus can precompute the key stream for
encryption and decryption. By default, CTR mode does not require
plaintext padding, however implementations MAY apply padding to the
packets. If the last key block is larger than the last plaintext block
the resulted value is truncated to the size of the plaintext block and
the most significant bits are used. When sending authentication data
inside packets the maximum amount of padding SHOULD be applied with
CTR mode as well.
In CTR mode only the encryption operation of the cipher is used. The
decryption operation is not needed since both encryption and decryption
process is simple XOR with the plaintext block and the key stream block.
The counter block is used to create the key for the CTR mode. When
SILC specifications refer to Initial Vector (IV) in general cases, in
case of CTR mode it refers to the counter block. The format of the
128 bit counter block is as follows:
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Truncated HASH from SKE |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Sending/Receiving IV from SKE |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Block Counter |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 6: Counter Block
Riikonen [Page 24]
Internet Draft 26 November 2002
o Truncated HASH from SKE (4 bytes) - This value is the first 4
bytes from the HASH value that was computed as a result of SKE
protocol. This acts as session identifier and each rekey MUST
produce a new HASH value.
o Sending/Receiving IV from SKE (8 bytes) - This value is the
first 8 bytes from the Sending IV or Receiving IV generated in
the SKE protocol. When this mode is used to encrypt sending
traffic the Sending IV is used, when used to decrypt receiving
traffic the Receiving IV is used. This assures that two parties
of the protocol use different IV for sending traffic. Each rekey
MUST produce a new value.
o Block Counter (4 bytes) - This is the counter value for the
counter block and is MSB ordered number starting from one (1)
value for first block and incrementing for subsequent blocks.
The same value MUST NOT be used twice. The rekey MUST be
performed before this counter value wraps.
CTR mode MUST NOT be used with "none" MAC. Implementations also MUST
assure that the same counter block is not used to encrypt more than
one block. Also, the key material used with CTR mode MUST be fresh
key material. Static keys (pre-shared keys) MUST NOT be used with
CTR mode. For this reason using CTR mode to encrypt for example
channel messages or private messages with a pre-shared key is
inappropriate. For private messages, the Key Agreement could be
performed to produce fresh key material.
If the IV Included flag was negotiated in SKE, implementations SHOULD
still use the same counter block format as defined above. However,
implementations are RECOMMENDED to replace the Truncated HASH field
with a 32 bit random value for each IV (counter block) per encrypted
SILC packet. Also note, that in this case the decryption process is
not stateful and receiver cannot precompute the key stream.
3.10.1.3 Randomized CBC Mode
The "rcbc" encryption mode is CBC mode with randomized IV. This means
that each IV for each packet MUST be chosen randomly. When encrypting
more than one block the normal inter-packet chaining is used, but for
the first block new random IV is selected in each packet. In this mode
the IV is appended at the end of the last ciphertext block and thus
delivered to the recipient. This mode increases the ciphertext size by
one ciphertext block. Note also that some data payloads in SILC are
capable of delivering the IV to the recipient. When explicitly
encrypting these payloads with randomized CBC the IV MUST NOT be appended
at the end of the ciphertext. When encrypting these payloads with
Riikonen [Page 25]
Internet Draft 26 November 2002
"cbc" mode they implicitly become randomized CBC since the IV is
usually selected random and included in the ciphertext. In these
cases using either CBC or randomized CBC is actually equivalent.
3.10.2 Public Key Algorithms
Public keys are used in SILC to authenticate entities in SILC network
and to perform other tasks related to public key cryptography. The
public keys are also used in the SILC Key Exchange protocol [SILC3].
The following public key algorithms are defined in SILC protocol:
rsa RSA (REQUIRED)
dss DSS (OPTIONAL)
DSS is described in [Menezes]. The RSA MUST be implemented according
PKCS #1 [PKCS1]. The mandatory PKCS #1 implementation in SILC MUST be
compliant to either PKCS #1 version 1.5 or newer with the following
notes: The signature encoding is always in same format as the encryption
encoding regardless of the PKCS #1 version. The signature with appendix
(with hash algorithm OID in the data) MUST NOT be used in the SILC. The
rationale for this is that there is no binding between the PKCS #1 OIDs
and the hash algorithms used in the SILC protocol. Hence, the encoding
is always in PKCS #1 version 1.5 format.
Additional public key algorithms MAY be defined to be used in SILC.
When signatures are computed in SILC the computing of the signature is
represented as sign(). The signature computing procedure is dependent
of the public key algorithm, and the public key or certificate encoding.
When using SILC public key the signature is computed as described in
previous paragraph for RSA and DSS keys. If the hash function is not
specified separately for signing process sha1 MUST be used. When using
SSH2 public keys the signature is computed as described in [SSH-TRANS].
When using X.509 version 3 certificates the signature is computed as
described in [PKCS7]. When using OpenPGP certificates the signature is
computed as described in [PGP].
3.10.3 Hash Functions
Hash functions are used as part of MAC algorithms defined in the next
section. They are also used in the SILC Key Exchange protocol defined
in the [SILC3].
The following Hash algorithm are defined in SILC protocol:
Riikonen [Page 26]
Internet Draft 26 November 2002
sha1 SHA-1, length = 20 (REQUIRED)
md5 MD5, length = 16 (RECOMMENDED)
3.10.4 MAC Algorithms
Data integrity is protected by computing a message authentication code
(MAC) of the packet data. See [SILC2] for details how to compute the
MAC.
The following MAC algorithms are defined in SILC protocol:
hmac-sha1-96 HMAC-SHA1, length = 12 bytes (REQUIRED)
hmac-md5-96 HMAC-MD5, length = 12 bytes (OPTIONAL)
hmac-sha1 HMAC-SHA1, length = 20 bytes (OPTIONAL)
hmac-md5 HMAC-MD5, length = 16 bytes (OPTIONAL)
none No MAC (OPTIONAL)
The "none" MAC is not recommended to be used as the packet is not
authenticated when MAC is not computed. It is recommended that no
client or server would accept none MAC except in special debugging
mode.
The HMAC algorithm is described in [HMAC] and hash algorithms that
are used as part of the HMACs are described in [Scheneir] and in
[Menezes].
Additional MAC algorithms MAY be defined to be used in SILC.
3.10.5 Compression Algorithms
SILC protocol supports compression that may be applied to unencrypted
data. It is recommended to use compression on slow links as it may
significantly speed up the data transmission. By default, SILC does not
use compression which is the mode that must be supported by all SILC
implementations.
The following compression algorithms are defined:
none No compression (REQUIRED)
zlib GNU ZLIB (LZ77) compression (OPTIONAL)
Additional compression algorithms MAY be defined to be used in SILC.
Riikonen [Page 27]
Internet Draft 26 November 2002
3.11 SILC Public Key
This section defines the type and format of the SILC public key. All
implementations MUST support this public key type. See [SILC3] for
other optional public key and certificate types allowed in the SILC
protocol. Public keys in SILC may be used to authenticate entities
and to perform other tasks related to public key cryptography.
The format of the SILC Public Key is as follows:
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Public Key Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Algorithm Name Length | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| |
~ Algorithm Name ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Identifier Length | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| |
~ Identifier ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ Public Data ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 5: SILC Public Key
o Public Key Length (4 bytes) - Indicates the full length
of the public key, not including this field.
o Algorithm Name Length (2 bytes) - Indicates the length
of the Algorithm Length field, not including this field.
o Algorithm name (variable length) - Indicates the name
of the public key algorithm that the key is. See the
section 3.10.2 Public Key Algorithms for defined names.
o Identifier Length (2 bytes) - Indicates the length of
the Identifier field, not including this field.
Riikonen [Page 28]
Internet Draft 26 November 2002
o Identifier (variable length) - Indicates the identifier
of the public key. This data can be used to identify
the owner of the key. The identifier is of the following
format:
UN User name
HN Host name or IP address
RN Real name
E EMail address
O Organization
C Country
Examples of an identifier:
`UN=priikone, HN=poseidon.pspt.fi, E=priikone@poseidon.pspt.fi'
`UN=sam, HN=dummy.fi, RN=Sammy Sam, O=Company XYZ, C=Finland'
At least user name (UN) and host name (HN) MUST be provided as
identifier. The fields are separated by commas (`,'). If
comma is in the identifier string it must be written as `\,',
for example, `O=Company XYZ\, Inc.'.
o Public Data (variable length) - Includes the actual
public data of the public key.
The format of this field for RSA algorithm is
as follows:
4 bytes Length of e
variable length e
4 bytes Length of n
variable length n
The format of this field for DSS algorithm is
as follows:
4 bytes Length of p
variable length p
4 bytes Length of q
variable length q
4 bytes Length of g
variable length g
4 bytes Length of y
variable length y
Riikonen [Page 29]
Internet Draft 26 November 2002
The variable length fields are multiple precession
integers encoded as strings in both examples.
Other algorithms must define their own type of this
field if they are used.
All fields in the public key are in MSB (most significant byte first)
order. All strings in the public key are UTF-8 encoded.
If an external protocol need to refer to SILC Public Key by name, the
name "silc-rsa" and "silc-dss" for SILC Public Key based on RSA algorithm
and SILC Public Key based on DSS algorithm, respectively, are to be used.
However, this SILC specification does not use these names directly, and
they are defined here for external protocols (protocols that may like
to use SILC Public Key).
3.12 SILC Version Detection
The version detection of both client and server is performed at the
connection phase while executing the SILC Key Exchange protocol. The
version identifier is exchanged between initiator and responder. The
version identifier is of the following format:
SILC-<protocol version>-<software version>
The version strings are of the following format:
protocol version = <major>.<minor>
software version = <major>[.<minor>[.<build or vendor string>]]
Protocol version MUST provide both major and minor version. Currently
implementations MUST set the protocol version and accept at least the
protocol version as SILC-1.2-<software version>. If new protocol version
causes incompatibilities with older version the <minor> version number
MUST be incremented. The <major> is incremented if new protocol version
is fully incompatible.
Software version MAY provide major, minor and build (vendor) version.
The software version MAY be freely set and accepted. The version string
MUST consist of printable US-ASCII characters.
Thus, the version strings could be, for example:
SILC-1.1-2.0.2
SILC-1.0-1.2
SILC-1.2-1.0.VendorXYZ
SILC-1.2-2.4.5 Vendor Limited
Riikonen [Page 30]
Internet Draft 26 November 2002
3.13 Backup Routers
Backup routers may exist in the cell in addition of the primary router.
However, they must not be active routers and act as routers in the cell.
Only one router may be acting as primary router in the cell. In the case
of failure of the primary router may one of the backup routers become
active. The purpose of backup routers are in case of failure of the
primary router to maintain working connections inside the cell and outside
the cell and to avoid netsplits.
Backup routers are normal servers in the cell that are prepared to take
over the tasks of the primary router if needed. They need to have at
least one direct and active connection to the primary router of the cell.
This communication channel is used to send the router information to
the backup router. When the backup router connects to the primary router
of the cell it MUST present itself as router server in the Connection
Authentication protocol, even though it is normal server as long as the
primary router is available. Reason for this is that the configuration
needed in the responder end requires usually router connection level
configuration. The responder, however must understand and treat the
connection as normal server (except when feeding router level data to
the backup router).
Backup router must know everything that the primary router knows to be
able to take over the tasks of the primary router. It is the primary
router's responsibility to feed the data to the backup router. If the
backup router does not know all the data in the case of failure some
connections may be lost. The primary router of the cell must consider
the backup router being actual router server when it feeds the data to
it.
In addition of having direct connection to the primary router of the
cell, the backup router must also have connection to the same router
the primary router of the cell is connected. However, it must not be
active router connection meaning that the backup router must not use
that channel as its primary route and it must not notify the router
about having connected servers, channels and clients behind it. It
merely connects to the router. This sort of connection is later
referred as being passive connection. Some keepalive actions may be
needed by the router to keep the connection alive.
It is required that other normal servers have passive connections to
the backup router(s) in the cell. Some keepalive actions may be needed
by the server to keep the connection alive. After they notice the
failure of the primary router they must start using the connection to
the first backup router as their primary route.
Also, if any other router in the network is using the cell's primary
Riikonen [Page 31]
Internet Draft 26 November 2002
router as its own primary router, it must also have passive connection
to the cell's backup router. It too is prepared to switch to use the
backup router as its new primary router as soon as the original primary
router becomes unresponsive.
All of the parties of this protocol knows which one is the backup router
of the cell from their local configuration. Each of the entity must
be configured accordingly and care must be taken when configuring the
backup routers, servers and other routers in the network.
It must be noted that some of the channel messages and private messages
may be lost during the switch to the backup router. The announcements
assures that the state of the network is not lost during the switch.
It is RECOMMENDED that there would be at least one backup router in
the cell. It is NOT RECOMMENDED to have all servers in the cell acting
as backup routers as it requires establishing several connections to
several servers in the cell. Large cells can easily have several
backup routers in the cell.
The order of the backup routers are decided at the configuration phase.
All the parties of this protocol must be configured accordingly to
understand the order of the backup routers. It is not required that
the backup server is actually active server in the cell. Backup router
may be a spare server in the cell that does not accept normal client
connections at all. It may be reserved purely for the backup purposes.
These, however, are cell management issues.
If also the first backup router is down as well and there is another
backup router in the cell then it will start acting as the primary
router as described above.
3.13.1 Switching to Backup Router
When the primary router of the cell becomes unresponsive, for example
by sending EOF to the connection, all the parties of this protocol MUST
replace the old connection to the primary router with first configured
backup router. The backup router usually needs to do local modifications
to its database in order to update all the information needed to maintain
working routes. The backup router must understand that clients that
were originated from the primary router are now originated from some of
the existing server connections and must update them accordingly. It
must also remove those clients that were owned by the primary router
since those connections were lost when the primary router became
unresponsive.
All the other parties of the protocol must also update their local
Riikonen [Page 32]
Internet Draft 26 November 2002
database to understand that the route to the primary router will now go
to the backup router.
Servers connected to the backup router MUST send SILC_PACKET_RESUME_ROUTER
packet with type number 21, to indicate that the server will start using
the backup router as primary router. The backup router MUST NOT allow
this action if it detects that primary is still up and running. If
backup router knows that primary is up and running it MUST send type
number 22 back to the server. The server then MUST NOT use the backup
as primary router, but must try to establish connection back to the
primary router. If the action is allowed type number 21 is sent back
to the server from the backup router.
The servers connected to the backup router must then announce their
clients, channels, channel users, channel user modes and channel modes
to the backup router. This is to assure that none of the important notify
packets were lost during the switch to the backup router. The backup
router must check which of these announced entities it already have
and distribute the new ones to the primary route.
The backup router too must announce its servers, clients, channels
and other information to the new primary router. The primary router
of the backup router too must announce its informations to the backup
router. Both must process only the ones they do not know about. If
any of the announced modes does not match then they are enforced in
normal manner defined later in this specification.
3.13.2 Resuming Primary Router
Usually the primary router is unresponsive only a short period of time
and it is intended that the original router of the cell will resume
its position as primary router when it comes back online. The backup
router that is now acting as primary router of the cell must constantly
try to connect to the original primary router of the cell. It is
RECOMMENDED that it would try to reconnect in 30 second intervals to
the primary router.
When the connection is established to the primary router the backup
resuming protocol is executed. The protocol is advanced as follows:
1. Backup router sends SILC_PACKET_RESUME_ROUTER packet with type
value 1 the primary router that came back online. The packet
will indicate the primary router has been replaced by the backup
router. After sending the packet the backup router will announce
all of its channels, channel users, modes etc. to the primary
router.
Riikonen [Page 33]
Internet Draft 26 November 2002
If the primary knows that it has not been replaced (for example
the backup itself disconnected from the primary router and thinks
that it is now primary in the cell) the primary router send
SILC_PACKET_FAILURE with the type value 1 back to the backup
router. If backup receives this it MUST NOT continue with the
backup resuming protocol.
2. Backup router sends SILC_PACKET_RESUME_ROUTER packet with type
value 2 to its current primary router to indicate that it will
resign as being primary router. Then, backup router sends the
SILC_PACKET_RESUME_ROUTER packet with type value 1 to all
connected servers to also indicate that it will resign as being
primary router.
3. Backup router also send SILC_PACKET_RESUME_ROUTER packet with
type value 2 to the router that is using the backup router
currently as its primary router.
4. Any server and router that receives the SILC_PACKET_RESUME_ROUTER
with type value 1 or 2 must reconnect immediately to the
primary router of the cell that came back online. After they
have created the connection they MUST NOT use that connection
as active primary route but still route all packets to the
backup router. After the connection is created they MUST send
SILC_PACKET_RESUME_ROUTER with type value 3 back to the
backup router. The session ID value found in the first packet
MUST be set in this packet.
5. Backup router MUST wait for all packets with type value 3 before
it continues with the protocol. It knows from the session ID values
set in the packet when it have received all packets. The session
value should be different in all packets it have sent earlier.
After the packets is received the backup router sends the
SILC_PACKET_RESUME_ROUTER packet with type value 4 to the
primary router that came back online. This packet will indicate
that the backup router is now ready to resign as being primary
router. The session ID value in this packet MUST be the same as
in first packet sent to the primary router. During this time
the backup router must still route all packets it is receiving
from server connections.
6. The primary router receives the packet and send the
SILC_PACKET_RESUME_ROUTER with type value 5 to all connected servers
including the backup router. It also sends the packet with type
value 6 to its primary router, and to the router that is using
it as its primary router. The Session ID value in this packet
SHOULD be zero (0).
Riikonen [Page 34]
Internet Draft 26 November 2002
7. Any server and router that receives the SILC_PACKET_RESUME_ROUTER
with type value 5 or 6 must switch their primary route to the
new primary router and remove the route for the backup router, since
it is not anymore the primary router of the cell. They must also
update their local database to understand that the clients are
not originated from the backup router but from the locally connected
servers. After that they MUST announce their channels, channel
users, modes etc. to the primary router. They must not use the
backup router connection after this and the connection is considered
to be passive connection. The implementations SHOULD be able
to disable the connection without closing the actual link.
After this protocol is executed the backup router is now again normal
server in the cell that has the backup link to the primary router. The
primary router feeds the router specific data again to the backup router.
All server connections in the backup router are considered passive
connections.
When the primary router of the cell comes back online and connects
to its primary router, the remote primary router must send the
SILC_PACKET_RESUME_ROUTER with type value 20 indicating that the
connection is not allowed since the router has been replaced by an
backup router. The session ID value in this packet SHOULD be zero (0).
When the router receives this packet it must not use the connection
as active connection but to understand that it cannot act as primary
router in the cell. It must wait that the backup router connects to
it, and the backup resuming protocol is executed.
The following type values has been defined for SILC_PACKET_RESUME_ROUTER
packet:
1 SILC_SERVER_BACKUP_START
2 SILC_SERVER_BACKUP_START_GLOBAL
3 SILC_SERVER_BACKUP_START_CONNECTED
4 SILC_SERVER_BACKUP_START_ENDING
5 SILC_SERVER_BACKUP_START_RESUMED
6 SILC_SERVER_BACKUP_START_RESUMED_GLOBAL
20 SILC_SERVER_BACKUP_START_REPLACED
21 SILC_SERVER_BACKUP_START_USE
22 SILC_SERVER_BACKUP_START_USE_DENIED
If any other value is found in the type field the packet must be
discarded. The SILC_PACKET_RESUME_ROUTER packet and its payload
is defined in [SILC2].
Riikonen [Page 35]
Internet Draft 26 November 2002
3.13.3 Discussion on Backup Router Scheme
It is clear that this backup router support is not able to handle all
possible situations arising in unreliable network environment. This
scheme for example does not handle situation when the router actually
does not go offline but the network link goes down temporarily. It would
require some intelligence to figure out when it is best time to switch
to the backup router. To make it even more complicated it is possible
that the backup router may have not lost the network link to the primary
router.
Other possible situation is when the network link is lost temporarily
between two primary routers in the SILC network. Unless the routers
notice the link going down they cannot perhaps find alternative routes.
Worst situation is when the link goes down only for a short period of
time, thus causing lag. Should the routers or servers find alternative
routes if they cannot get response from the router during the lag?
When alternative routes are being found it must be careful not to
mess up existing primary routes between routers in the network.
It is suggested that the current backup router scheme is only temporary
solution and existing backup router protocols are studied further. It
is also suggested that the backup router specification will be separated
from this SILC specification Internet-Draft and additional specification
is written on the subject.
4 SILC Procedures
This section describes various SILC procedures such as how the
connections are created and registered, how channels are created and
so on. The section describes the procedures only generally as details
are described in [SILC2] and [SILC3].
4.1 Creating Client Connection
This section describes the procedure when client connects to SILC server.
When client connects to server the server MUST perform IP address lookup
and reverse IP address lookup to assure that the origin host really is
who it claims to be. Client, host, connecting to server SHOULD have
both valid IP address and fully qualified domain name (FQDN).
After that the client and server performs SILC Key Exchange protocol
which will provide the key material used later in the communication.
The key exchange protocol MUST be completed successfully before the
connection registration may continue. The SILC Key Exchange protocol
is described in [SILC3].
Riikonen [Page 36]
Internet Draft 26 November 2002
Typical server implementation would keep a list of connections that it
allows to connect to the server. The implementation would check, for
example, the connecting client's IP address from the connection list
before the SILC Key Exchange protocol has been started. Reason for
this is that if the host is not allowed to connect to the server there
is no reason to perform the key exchange protocol.
After successful key exchange protocol the client and server performs
connection authentication protocol. The purpose of the protocol is to
authenticate the client connecting to the server. Flexible
implementation could also accept the client to connect to the server
without explicit authentication. However, if authentication is
desired for a specific client it may be based on passphrase or
public key authentication. If authentication fails the connection
MUST be terminated. The connection authentication protocol is described
in [SILC3].
After successful key exchange and authentication protocol the client
registers itself by sending SILC_PACKET_NEW_CLIENT packet to the
server. This packet includes various information about the client
that the server uses to create the client. Server creates the client
and sends SILC_PACKET_NEW_ID to the client which includes the created
Client ID that the client MUST start using after that. After that
all SILC packets from the client MUST have the Client ID as the
Source ID in the SILC Packet Header, described in [SILC2].
Client MUST also get the server's Server ID that is to be used as
Destination ID in the SILC Packet Header when communicating with
the server (for example when sending commands to the server). The
ID may be resolved in two ways. Client can take the ID from an
previously received packet from server that MUST include the ID,
or to send SILC_COMMAND_INFO command and receive the Server ID as
command reply.
Server MAY choose not to use the information received in the
SILC_PACKET_NEW_CLIENT packet. For example, if public key or
certificate were used in the authentication, server MAY use those
informations rather than what it received from client. This is suitable
way to get the true information about client if it is available.
The nickname of client is initially set to the username sent in the
SILC_PACKET_NEW_CLIENT packet. User should set the nickname to more
suitable by sending SILC_COMMAND_NICK command. However, this is not
required as part of registration process.
Server MUST also distribute the information about newly registered
client to its router (or if the server is router, to all routers in
the SILC network). More information about this in [SILC2].
Riikonen [Page 37]
Internet Draft 26 November 2002
Router server MUST also check whether some client in the local cell
is watching for the nickname this new client has, and send the
SILC_NOTIFY_TYPE_WATCH to the watcher.
4.2 Creating Server Connection
This section describes the procedure when server connects to its
router (or when router connects to other router, the cases are
equivalent). The procedure is very much alike when client connects
to the server thus it is not repeated here.
One difference is that server MUST perform connection authentication
protocol with proper authentication. A proper authentication is based
on passphrase authentication or public key authentication based on
digital signatures.
After server and router has successfully performed the key exchange
and connection authentication protocol, the server register itself
to the router by sending SILC_PACKET_NEW_SERVER packet. This packet
includes the server's Server ID that it has created by itself and
other relevant information about the server.
After router has received the SILC_PACKET_NEW_SERVER packet it
distributes the information about newly registered server to all routers
in the SILC network. More information about this in [SILC2].
As client needed to resolve the destination ID this MUST be done by the
server that connected to the router, as well. The way to resolve it is
to get the ID from previously received packet. The server MAY also
use SILC_COMMAND_INFO command to resolve the ID. Server MUST also start
using its own Server ID as Source ID in SILC Packet Header and the
router's Server ID as Destination when communicating with the router.
4.2.1 Announcing Clients, Channels and Servers
After server or router has connected to the remote router, and it already
has connected clients and channels it MUST announce them to the router.
If the server is router server, also all the local servers in the cell
MUST be announced.
All clients are announced by compiling a list of ID Payloads into the
SILC_PACKET_NEW_ID packet. All channels are announced by compiling a
list of Channel Payloads into the SILC_PACKET_NEW_CHANNEL packet.
Channels' mode and founder public key and other channel mode specific
data is announced by sending SILC_NOTIFY_TYPE_CMODE_CHANGE notify list.
Also, the channel users on the channels must be announced by compiling a
Riikonen [Page 38]
Internet Draft 26 November 2002
list of Notify Payloads with the SILC_NOTIFY_TYPE_JOIN notify type into
the SILC_PACKET_NOTIFY packet. The users' modes on the channel must
also be announced by compiling list of Notify Payloads with the
SILC_NOTIFY_TYPE_CUMODE_CHANGE notify type into the SILC_PACKET_NOTIFY
packet.
The router MUST also announce the local servers by compiling list of
ID Payloads into the SILC_PACKET_NEW_ID packet.
Also, clients' modes (user modes in SILC) MUST be announced. This is
done by compiling a list of Notify Payloads with SILC_NOTIFY_UMODE_CHANGE
notify type into the SILC_PACKET_NOTIFY packet. Also, channel's topics
MUST be announced by compiling a list of Notify Payloads with the
SILC_NOTIFY_TOPIC_SET notify type into the SILC_PACKET_NOTIFY packet.
The router which receives these lists MUST process them and broadcast
the packets to its primary route. When processing the announced channels
and channel users the router MUST check whether a channel exists already
with the same name. If channel exists with the same name it MUST check
whether the Channel ID is different. If the Channel ID is different the
router MUST send the notify type SILC_NOTIFY_TYPE_CHANNEL_CHANGE to the
server to force the channel ID change to the ID the router has. If the
mode of the channel is different the router MUST send the notify type
SILC_NOTIFY_TYPE_CMODE_CHANGE to the server to force the mode change
to the mode that the router has.
The router MUST also generate new channel key and distribute it to the
channel. The key MUST NOT be generated if the SILC_CMODE_PRIVKEY mode
is set.
If the channel has channel founder on the router the router MUST send
the notify type SILC_NOTIFY_TYPE_CUMODE_CHANGE to the server to force
the mode change for the channel founder on the server. The channel
founder privileges MUST be removed.
The router processing the channels MUST also compile a list of
Notify Payloads with the SILC_NOTIFY_TYPE_JOIN notify type into the
SILC_PACKET_NOTIFY and send the packet to the server. This way the
server (or router) will receive the clients on the channel that
the router has.
4.3 Joining to a Channel
This section describes the procedure when client joins to a channel.
Client joins to channel by sending command SILC_COMMAND_JOIN to the
server. If the receiver receiving join command is normal server the
server MUST check its local list whether this channel already exists
Riikonen [Page 39]
Internet Draft 26 November 2002
locally. This would indicate that some client connected to the server
has already joined to the channel. If this is case the client is
joined to the channel, new channel key is created and information about
newly joined channel is sent to the router. The router is informed
by sending SILC_NOTIFY_TYPE_JOIN notify type. The notify type MUST
also be sent to the local clients on the channel. The new channel key
is also sent to the router and to local clients on the channel.
If the channel does not exist in the local list the client's command
MUST be sent to the router which will then perform the actual joining
procedure. When server receives the reply to the command from the
router it MUST be sent to the client which sent the command originally.
Server will also receive the channel key from the server that it MUST
send to the client which originally requested the join command. The
server MUST also save the channel key.
If the receiver of the join command is router it MUST first check its
local list whether anyone in the cell has already joined to the channel.
If this is the case the client is joined to the channel and reply is
sent to the client. If the command was sent by server the command reply
is sent to the server which sent it. Then the router MUST also create
new channel key and distribute it to all clients on the channel and
all servers that has clients on the channel. Router MUST also send
the SILC_NOTIFY_TYPE_JOIN notify type to local clients on the channel
and to local servers that has clients on the channel.
If the channel does not exist on the router's local list it MUST
check the global list whether the channel exists at all. If it does
the client is joined to the channel as described previously. If
the channel does not exist the channel is created and the client
is joined to the channel. The channel key is also created and
distributed as previously described. The client joining to the created
channel is made automatically channel founder and both channel founder
and channel operator privileges is set for the client.
If the router created the channel in the process, information about the
new channel MUST be broadcasted to all routers. This is done by
broadcasting SILC_PACKET_NEW_CHANNEL packet to the router's primary
route. When the router joins the client to the channel it MUST also
send information about newly joined client to all routers in the SILC
network. This is done by broadcasting the SILC_NOTIFY_TYPE_JOIN notify
type to the router's primary route.
It is important to note that new channel key is created always when
new client joins to channel, whether the channel has existed previously
or not. This way the new client on the channel is not able to decrypt
any of the old traffic on the channel. Client which receives the reply to
the join command MUST start using the received Channel ID in the channel
Riikonen [Page 40]
Internet Draft 26 November 2002
message communication thereafter. Client also receives the key for the
channel in the command reply. Note that the channel key is never
generated if the SILC_CMODE_PRIVKEY mode is set.
4.4 Channel Key Generation
Channel keys are created by router which creates the channel by taking
enough randomness from cryptographically strong random number generator.
The key is generated always when channel is created, when new client
joins a channel and after the key has expired. Key could expire for
example in an hour.
The key MUST also be re-generated whenever some client leaves a channel.
In this case the key is created from scratch by taking enough randomness
from the random number generator. After that the key is distributed to
all clients on the channel. However, channel keys are cell specific thus
the key is created only on the cell where the client, which left the
channel, exists. While the server or router is creating the new channel
key, no other client may join to the channel. Messages that are sent
while creating the new key are still processed with the old key. After
server has sent the SILC_PACKET_CHANNEL_KEY packet MUST client start
using the new key. If server creates the new key the server MUST also
send the new key to its router. See [SILC2] on more information about
how channel messages must be encrypted and decrypted when router is
processing them.
When client receives the SILC_PACKET_CHANNEL_KEY packet with the
Channel Key Payload it MUST process the key data to create encryption
and decryption key, and to create the HMAC key that is used to compute
the MACs of the channel messages. The processing is as follows:
channel_key = raw key data
HMAC key = hash(raw key data)
The raw key data is the key data received in the Channel Key Payload.
The hash() function is the hash function used in the HMAC of the channel.
Note that the server also MUST save the channel key.
4.5 Private Message Sending and Reception
Private messages are sent point to point. Client explicitly destine
a private message to specific client that is delivered to only to that
client. No other client may receive the private message. The receiver
of the private message is destined in the SILC Packet Header as any
other packet as well.
Riikonen [Page 41]
Internet Draft 26 November 2002
If the sender of a private message does not know the receiver's Client
ID, it MUST resolve it from server. There are two ways to resolve the
client ID from server; it is RECOMMENDED that client implementations
send SILC_COMMAND_IDENTIFY command to receive the Client ID. Client
MAY also send SILC_COMMAND_WHOIS command to receive the Client ID.
If the sender has received earlier a private message from the receiver
it should have cached the Client ID from the SILC Packet Header.
If server receives a private message packet which includes invalid
destination Client ID the server MUST send SILC_NOTIFY_TYPE_ERROR
notify to the client with error status indicating that such Client ID
does not exist.
See [SILC2] for description of private message encryption and decryption
process.
4.6 Private Message Key Generation
Private message MAY be protected with a key generated by the client.
The key may be generated and sent to the other client by sending packet
SILC_PACKET_PRIVATE_MESSAGE_KEY which travels through the network
and is secured by session keys. After that the private message key
is used in the private message communication between those clients.
The key sent inside the payload SHOULD be randomly generated. This
packet MUST NOT be used to send pre-shared keys.
Other choice is to entirely use keys that are not sent through
the SILC network at all. This significantly adds security. This key
could be a pre-shared-key that is known by both of the clients. Both
agree about using the key and starts sending packets that indicate
that the private message is secured using private message key. In
case of pre-shared keys (static keys) the IV used in encryption SHOULD
be chosen randomly.
It is also possible to negotiate fresh key material by performing
Key Agreement. The SILC_PACKET_KEY_AGREEMENT packet MAY be used to
negotiate the fresh key material. In this case the resulted key
material is used to secure the private messages. Also, the IV used
in encryption is used as defined in [SILC3], unless otherwise stated
by the encryption mode used. By performing Key Agreement the clients
may negotiate the cipher and HMAC to be used in the private message
encryption and to negotiate additional security parameters.
If the key is pre-shared key or other key material not generated by
Key Agreement, then the key material SHOULD be processed as defined
in [SILC3]. In the processing, however, the HASH, as defined in
[SILC3] MUST be ignored. After processing the key material it is
Riikonen [Page 42]
Internet Draft 26 November 2002
employed as defined in [SILC3]. In this case also, implementations
SHOULD use the SILC protocol's mandatory cipher and HMAC in private
message encryption.
4.7 Channel Message Sending and Reception
Channel messages are delivered to group of users. The group forms a
channel and all clients on the channel receives messages sent to the
channel.
Channel messages are destined to channel by specifying the Channel ID
as Destination ID in the SILC Packet Header. The server MUST then
distribute the message to all clients on the channel by sending the
channel message destined explicitly to a client on the channel.
If server receives a channel message packet which includes invalid
destination Channel ID the server MUST send SILC_NOTIFY_TYPE_ERROR
notify to the sender with error status indicating that such Channel ID
does not exist.
See the [SILC2] for description of channel message routing for router
servers, and channel message encryption and decryption process.
4.8 Session Key Regeneration
Session keys MUST be regenerated periodically, say, once in an hour.
The re-key process is started by sending SILC_PACKET_REKEY packet to
other end, to indicate that re-key must be performed. The initiator
of the connection SHOULD initiate the re-key.
If perfect forward secrecy (PFS) flag was selected in the SILC Key
Exchange protocol [SILC3] the re-key MUST cause new key exchange with
SKE protocol. In this case the protocol is secured with the old key
and the protocol results to new key material. See [SILC3] for more
information. After the SILC_PACKET_REKEY packet is sent the sender
will perform the SKE protocol.
If PFS flag was set the resulted key material is processed as described
in the section Processing the Key Material in [SILC3]. The difference
with re-key in the processing is that the initial data for the hash
function is just the resulted key material and not the HASH as it
is not computed at all with re-key. Other than that, the key processing
it equivalent to normal SKE negotiation.
If PFS flag was not set, which is the default case, then re-key is done
without executing SKE protocol. In this case, the new key is created by
Riikonen [Page 43]
Internet Draft 26 November 2002
providing the current sending encryption key to the SKE protocol's key
processing function. The process is described in the section Processing
the Key Material in [SILC3]. The difference in the processing is that
the initial data for the hash function is the current sending encryption
key and not the SKE's KEY and HASH values. Other than that, the key
processing is equivalent to normal SKE negotiation.
After both parties has regenerated the session key, both MUST send
SILC_PACKET_REKEY_DONE packet to each other. These packets are still
secured with the old key. After these packets, the subsequent packets
MUST be protected with the new key.
4.9 Command Sending and Reception
Client usually sends the commands in the SILC network. In this case
the client simply sends the command packet to server and the server
processes it and replies with command reply packet. See the [SILC3]
for detailed description of all commands.
However, if the server is not able to process the command, it is sent
to the server's router. This is case for example with commands such
as, SILC_COMMAND_JOIN and SILC_COMMAND_WHOIS commands. However, there
are other commands as well. For example, if client sends the WHOIS
command requesting specific information about some client the server must
send the WHOIS command to router so that all clients in SILC network
are searched. The router, on the other hand, sends the WHOIS command
further to receive the exact information about the requested client.
The WHOIS command travels all the way to the server which owns the client
and it replies with command reply packet. Finally, the server which
sent the command receives the command reply and it must be able to
determine which client sent the original command. The server then
sends command reply to the client. Implementations should have some
kind of cache to handle, for example, WHOIS information. Servers
and routers along the route could all cache the information for faster
referencing in the future.
The commands sent by server may be sent hop by hop until someone is able
to process the command. However, it is preferred to destine the command
as precisely as it is possible. In this case, other routers en route
MUST route the command packet by checking the true sender and true
destination of the packet. However, servers and routers MUST NOT route
command reply packets to clients coming from other server. Client
MUST NOT accept command reply packet originated from anyone else but
from its own server.
Riikonen [Page 44]
Internet Draft 26 November 2002
4.10 Closing Connection
When remote client connection is closed the server MUST send the notify
type SILC_NOTIFY_TYPE_SIGNOFF to its primary router and to all channels
the client was joined. The server MUST also save the client's information
for a period of time for history purposes.
When remote server or router connection is closed the server or router
MUST also remove all the clients that was behind the server or router
from the SILC Network. The server or router MUST also send the notify
type SILC_NOTIFY_TYPE_SERVER_SIGNOFF to its primary router and to all
local clients that are joined on the same channels with the remote
server's or router's clients.
Router server MUST also check whether some client in the local cell
is watching for the nickname this client has, and send the
SILC_NOTIFY_TYPE_WATCH to the watcher, unless the client which left
the network has the SILC_UMODE_REJECT_WATCHING user mode set.
4.11 Detaching and Resuming a Session
SILC protocol provides a possibility for a client to detach itself from
the network without actually signing off from the network. The client
connection to the server is closed but the client remains as valid client
in the network. The client may then later resume its session back from
any server in the network.
When client wishes to detach from the network it MUST send the
SILC_COMMAND_DETACH command to its server. The server then MUST set
SILC_UMODE_DETACHED mode to the client and send SILC_NOTIFY_UMODE_CHANGE
notify to its primary router, which will then MUST broadcast it further
to other routers in the network. This user mode indicates that the
client is detached from the network. Implementations MUST NOT use
the SILC_UMODE_DETACHED flag to determine whether a packet can be sent
to the client. All packets MUST still be sent to the client even if
client is detached from the network. Only the server that originally
had the active client connection is able to make the decision after it
notices that the network connection is not active. In this case the
default case is to discard the packet.
The SILC_UMODE_DETACHED flag cannot be set by client itself directly
with SILC_COMMAND_UMODE command, but only implicitly by sending the
SILC_COMMAND_DETACH command. The flag also cannot be unset by the
client, server or router with SILC_COMMAND_UMODE command, but only
implicitly by sending and receiving the SILC_PACKET_RESUME_CLIENT
packet.
Riikonen [Page 45]
Internet Draft 26 November 2002
When the client wishes to resume its session in the SILC Network it
connects to a server in the network, which MAY also be a different
from the original server, and performs normal procedures regarding
creating a connection as described in section 4.1. After the SKE
and the Connection Authentication protocols has been successfully
completed the client MUST NOT send SILC_PACKET_NEW_CLIENT packet, but
MUST send SILC_PACKET_RESUME_CLIENT packet. This packet is used to
perform the resuming procedure. The packet MUST include the detached
client's Client ID, which the client must know. It also includes
Authentication Payload which includes signature made with the client's
private key. The signature is computed as defined in the section
3.9.1. Thus, the authentication method MUST be based in public key
authentication.
When server receives the SILC_PACKET_RESUME_CLIENT packet it MUST
do the following: Server checks that the Client ID is valid client
and that it has the SILC_UMODE_DETACHED mode set. Then it verifies
the Authentication Payload with the detached client's public key.
If it does not have the public key it retrieves it by sending
SILC_COMMAND_GETKEY command to the server that has the public key from
the original client connection. The server MUST NOT use the public
key received in the SKE protocol for this connection. If the
signature is valid the server unsets the SILC_UMODE_DETACHED flag,
and sends the SILC_PACKET_RESUME_CLIENT packet to its primary router.
The routers MUST broadcast the packet and unset the SILC_UMODE_DETACHED
flag when the packet is received. If the server is router server it
also MUST send the SILC_PACKET_RESUME_CLIENT packet to the original
server whom owned the detached client.
The servers and routers that receives the SILC_PACKET_RESUME_CLIENT
packet MUST know whether the packet already has been received for
the client. It is protocol error to attempt to resume the client
session from more than one server. The implementations could set
internal flag that indicates that the client is resumed. If router
receive SILC_PACKET_RESUME_CLIENT packet for client that is already
resumed the client MUST be killed from the network. This would
indicate that the client is attempting to resume the session more
than once which is protocol error. In this case the router sends
SILC_NOTIFY_TYPE_KILLED to the client. All routers that detect
the same situation MUST also send the notify for the client.
The servers and routers that receive the SILC_PACKET_RESUME_CLIENT
must also understand that the client may not be found behind the
same server that it originally came from. They must update their
caches according this. The server that now owns the client session
MUST check whether the Client ID of the resumed client is based
on the server's Server ID. If it is not it creates a new Client
ID and send SILC_NOTIFY_TYPE_NICK_CHANGE to the network. It MUST
Riikonen [Page 46]
Internet Draft 26 November 2002
also send the channel keys of all channels that the client is
joined to the client since it does not have them. Whether the
Client ID was changed or not the server MUST send SILC_PACKET_NEW_ID
packet to the client. Only after this the client is resumed back
to the network and may start sending packets and messages.
It is also possible that the server does not know about the channels
that the client has joined. In this case it join the client internally
to the channels, generate new channel keys and distribute the keys
to the channels as described in section 4.4.
It is implementation issue for how long servers keep detached client
sessions. It is RECOMMENDED that the detached sessions would be
persistent as long as the server is running.
5 Security Considerations
Security is central to the design of this protocol, and these security
considerations permeate the specification. Common security considerations
such as keeping private keys truly private and using adequate lengths for
symmetric and asymmetric keys must be followed in order to maintain the
security of this protocol.
Special attention must also be paid on the servers and routers that are
running the SILC service. The SILC protocol's security depends greatly
on the security and the integrity of the servers and administrators that
are running the service. It is recommended that some form of registration
is required by the server and router administrator prior acceptance to
the SILC Network. Even though, the SILC protocol is secure in a network
of mutual distrust between clients, servers, routers and administrators
of the servers, the client should be able to trust the servers they are
using if they wish to do so.
It however must be noted that if the client requires absolute security
by not trusting any of the servers or routers in the SILC Network, it can
be accomplished by negotiating private keys outside the SILC Network,
either using SKE or some other key exchange protocol, or to use some
other external means for distributing the keys. This applies for all
messages, private messages and channel messages.
It is important to note that SILC, like any other security protocol is
not full proof system; the SILC servers and routers could very well be
compromised. However, to provide acceptable level of security and
usability for end user the protocol use many times session keys or other
keys generated by the servers to secure the messages. This is
intentional design feature to allow ease of use for end user. This way
the network is still usable, and remains encrypted even if the external
Riikonen [Page 47]
Internet Draft 26 November 2002
means of distributing the keys is not working. The implementation,
however, may like to not follow this design feature, and always negotiate
the keys outside SILC network. This is acceptable solution and many times
recommended. The implementation still must be able to work with the
server generated keys.
If this is unacceptable for the client or end user, the private keys
negotiated outside the SILC Network should always be used. In the end
it is always implementor's choice whether to negotiate private keys by
default or whether to use the keys generated by the servers.
It is also recommended that router operators in the SILC Network would
form a joint forum to discuss the router and SILC Network management
issues. Also, router operators along with the cell's server operators
should have a forum to discuss the cell management issues.
6 References
[SILC2] Riikonen, P., "SILC Packet Protocol", Internet Draft,
May 2002.
[SILC3] Riikonen, P., "SILC Key Exchange and Authentication
Protocols", Internet Draft, May 2002.
[SILC4] Riikonen, P., "SILC Commands", Internet Draft, May 2002.
[IRC] Oikarinen, J., and Reed D., "Internet Relay Chat Protocol",
RFC 1459, May 1993.
[IRC-ARCH] Kalt, C., "Internet Relay Chat: Architecture", RFC 2810,
April 2000.
[IRC-CHAN] Kalt, C., "Internet Relay Chat: Channel Management", RFC
2811, April 2000.
[IRC-CLIENT] Kalt, C., "Internet Relay Chat: Client Protocol", RFC
2812, April 2000.
[IRC-SERVER] Kalt, C., "Internet Relay Chat: Server Protocol", RFC
2813, April 2000.
[SSH-TRANS] Ylonen, T., et al, "SSH Transport Layer Protocol",
Internet Draft.
[PGP] Callas, J., et al, "OpenPGP Message Format", RFC 2440,
November 1998.
Riikonen [Page 48]
Internet Draft 26 November 2002
[SPKI] Ellison C., et al, "SPKI Certificate Theory", RFC 2693,
September 1999.
[PKIX-Part1] Housley, R., et al, "Internet X.509 Public Key
Infrastructure, Certificate and CRL Profile", RFC 2459,
January 1999.
[Schneier] Schneier, B., "Applied Cryptography Second Edition",
John Wiley & Sons, New York, NY, 1996.
[Menezes] Menezes, A., et al, "Handbook of Applied Cryptography",
CRC Press 1997.
[OAKLEY] Orman, H., "The OAKLEY Key Determination Protocol",
RFC 2412, November 1998.
[ISAKMP] Maughan D., et al, "Internet Security Association and
Key Management Protocol (ISAKMP)", RFC 2408, November
1998.
[IKE] Harkins D., and Carrel D., "The Internet Key Exchange
(IKE)", RFC 2409, November 1998.
[HMAC] Krawczyk, H., "HMAC: Keyed-Hashing for Message
Authentication", RFC 2104, February 1997.
[PKCS1] Kalinski, B., and Staddon, J., "PKCS #1 RSA Cryptography
Specifications, Version 2.0", RFC 2437, October 1998.
[RFC2119] Bradner, S., "Key Words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC2279] Yergeau, F., "UTF-8, a transformation format of ISO
10646", RFC 2279, January 1998.
[PKCS7] Kalinski, B., "PKCS #7: Cryptographic Message Syntax,
Version 1.5", RFC 2315, March 1998.
7 Author's Address
Pekka Riikonen
Snellmaninkatu 34 A 15
70100 Kuopio
Finland
EMail: priikone@iki.fi
Riikonen [Page 49]
Internet Draft 26 November 2002
This Internet-Draft expires 26 April 2003
Riikonen [Page 50]
