TLS-Session-Bound Access Tokens for OAuth 2.0
draft-mw-oauth-tls-session-bound-tokens-05
This document is an Internet-Draft (I-D).
Anyone may submit an I-D to the IETF.
This I-D is not endorsed by the IETF and has no formal standing in the
IETF standards process.
The information below is for an old version of the document.
| Document | Type |
This is an older version of an Internet-Draft whose latest revision state is "Active".
|
|
|---|---|---|---|
| Authors | Ramki Krishnan , A Prasad , Diego Lopez , Srinivasa Addepalli | ||
| Last updated | 2026-05-23 (Latest revision 2026-04-09) | ||
| RFC stream | (None) | ||
| Formats | |||
| Stream | Stream state | (No stream defined) | |
| Consensus boilerplate | Unknown | ||
| RFC Editor Note | (None) | ||
| IESG | IESG state | I-D Exists | |
| Telechat date | (None) | ||
| Responsible AD | (None) | ||
| Send notices to | (None) |
draft-mw-oauth-tls-session-bound-tokens-05
OAuth R. Krishnan
Internet-Draft JPMorgan Chase & Co
Intended status: Standards Track A. Prasad
Expires: 24 November 2026 Oracle
D. Lopez
Telefonica
S. Addepalli
Aryaka
23 May 2026
TLS-Session-Bound Access Tokens for OAuth 2.0
draft-mw-oauth-tls-session-bound-tokens-05
Abstract
This document defines a mechanism for binding OAuth 2.0 access tokens
to a specific mutual TLS (mTLS) connection. The binding is achieved
through a proof token that incorporates the TLS Exporter value
[RFC5705] derived from the current connection and an access token
hash, signed by the client's private key corresponding to its mTLS
certificate. This mechanism prevents stolen bearer tokens from being
replayed on a different TLS connection. The proof is constructed
once per (token, connection) pair and reused across all requests on
that connection, delivering session binding with no per-request
signing overhead and no additional key management beyond what mTLS
already provides. The mechanism is applicable to TLS 1.2, TLS 1.3,
and QUIC transports. While applicable to any OAuth 2.0 access token
presented over mTLS, this specification is primarily motivated by the
Token Exchange protocol [RFC8693], where multi-hop delegation chains
in autonomous, agent-driven architectures create elevated replay
risk.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
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."
Krishnan, et al. Expires 24 November 2026 [Page 1]
Internet-Draft TLS-Session-Bound-Tokens May 2026
This Internet-Draft will expire on 24 November 2026.
Copyright Notice
Copyright (c) 2026 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (https://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. The Bearer Token Replay Problem . . . . . . . . . . . . . 3
1.2. The Agentic AI Amplifier . . . . . . . . . . . . . . . . 4
1.3. Conventions and Terminology . . . . . . . . . . . . . . . 5
2. TLS Session Binding for Access Tokens . . . . . . . . . . . . 6
2.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2. TLS Exporter Derivation . . . . . . . . . . . . . . . . . 6
2.3. Session-Binding Proof Format . . . . . . . . . . . . . . 7
2.3.1. Header . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.2. Payload . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.3. Signature . . . . . . . . . . . . . . . . . . . . . . 9
2.4. Token Confirmation Claim . . . . . . . . . . . . . . . . 9
3. Protocol Flow . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1. Token Issuance with Session Binding . . . . . . . . . . . 9
3.2. Authorization Server Behavior . . . . . . . . . . . . . . 10
3.3. Resource Access with Session-Binding Proof . . . . . . . 10
3.3.1. Performance Characteristics . . . . . . . . . . . . . 12
3.3.2. Mid-Session Token Binding . . . . . . . . . . . . . . 12
3.4. Token Introspection Considerations . . . . . . . . . . . 14
3.5. Error Responses . . . . . . . . . . . . . . . . . . . . . 14
4. Integration with Existing Mechanisms . . . . . . . . . . . . 15
4.1. Relationship to RFC 8705 (mTLS-Bound Tokens) . . . . . . 15
4.2. Relationship to RFC 9449 (DPoP) . . . . . . . . . . . . . 15
4.3. Relationship to Token Binding (RFC 8471-8473) . . . . . . 16
4.4. Relationship to WIMSE WIT/WPT . . . . . . . . . . . . . . 17
4.5. Relationship to Transitive Attestation . . . . . . . . . 18
4.6. Relationship to RFC 8693 (Token Exchange) . . . . . . . . 18
5. Deployment Scope . . . . . . . . . . . . . . . . . . . . . . 19
5.1. Co-Located TLS Termination (Supported) . . . . . . . . . 19
5.2. Remote TLS Termination (Out of Scope) . . . . . . . . . . 19
Krishnan, et al. Expires 24 November 2026 [Page 2]
Internet-Draft TLS-Session-Bound-Tokens May 2026
5.3. QUIC and HTTP/3 . . . . . . . . . . . . . . . . . . . . . 20
6. Security Considerations . . . . . . . . . . . . . . . . . . . 20
6.1. Threat Model . . . . . . . . . . . . . . . . . . . . . . 20
6.1.1. T1: Cross-Connection Token Replay . . . . . . . . . . 20
6.1.2. T2: Cross-Host Token Replay . . . . . . . . . . . . . 21
6.1.3. T3: Token Exfiltration via LLM Prompt Injection . . . 21
6.1.4. T4: Token + Proof Exfiltration (Both Stolen) . . . . 21
6.1.5. T5: Multi-Hop Delegation Chain Compromise . . . . . . 22
6.2. Residual Risks . . . . . . . . . . . . . . . . . . . . . 22
6.2.1. Intra-Connection Replay . . . . . . . . . . . . . . . 22
6.2.2. Compromised Private Key . . . . . . . . . . . . . . . 23
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 23
7.1. OAuth Token Confirmation Methods . . . . . . . . . . . . 23
7.2. OAuth Dynamic Client Registration Metadata . . . . . . . 23
7.3. HTTP Header Fields . . . . . . . . . . . . . . . . . . . 24
7.3.1. Session-Binding-Proof . . . . . . . . . . . . . . . . 24
7.4. TLS Exporter Label . . . . . . . . . . . . . . . . . . . 24
8. Normative References . . . . . . . . . . . . . . . . . . . . 24
9. Informative References . . . . . . . . . . . . . . . . . . . 25
Appendix A. Contributors . . . . . . . . . . . . . . . . . . . . 26
Appendix B. Sidecar Deployment for Agentic AI . . . . . . . . . 26
B.1. Architecture . . . . . . . . . . . . . . . . . . . . . . 26
B.2. Request Flow . . . . . . . . . . . . . . . . . . . . . . 27
B.3. HTTP/2 Multi-User Multiplexing . . . . . . . . . . . . . 29
B.4. Security Benefits . . . . . . . . . . . . . . . . . . . . 29
B.5. Relationship to Existing Infrastructure . . . . . . . . . 30
B.6. Implementation Considerations . . . . . . . . . . . . . . 30
B.6.1. Connection Lifecycle Management . . . . . . . . . . . 30
Appendix C. SPIFFE/SPIRE Integration . . . . . . . . . . . . . . 31
C.1. Identity-Provenance Property . . . . . . . . . . . . . . 31
C.2. Deployment Variants . . . . . . . . . . . . . . . . . . . 32
C.2.1. Without Sidecar (Direct Integration) . . . . . . . . 32
C.2.2. With Sidecar (Recommended for Agentic AI) . . . . . . 34
C.3. Relationship to WIMSE WIT/WPT . . . . . . . . . . . . . . 35
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 35
1. Introduction
1.1. The Bearer Token Replay Problem
OAuth 2.0 access tokens are typically *bearer tokens*: any party in
possession of the token can use it to access protected resources,
regardless of the presenter's identity or the communication channel.
This is a known risk addressed by the OAuth 2.0 Security Best Current
Practice [I-D.ietf-oauth-security-topics].
Krishnan, et al. Expires 24 November 2026 [Page 3]
Internet-Draft TLS-Session-Bound-Tokens May 2026
The Token Exchange protocol [RFC8693] amplifies this risk by enabling
chained delegation across service boundaries. Each exchange produces
a new bearer token, and a compromise at any point in the chain
exposes downstream tokens.
Existing mitigations address parts of this problem:
* *RFC 8705 (mTLS Certificate-Bound Tokens)* [RFC8705]: Binds the
token to the client's X.509 certificate thumbprint. However, the
binding is to the _certificate identity_, not the _TLS
connection_. If the same certificate is used across connections,
or if the certificate and token are both exfiltrated, the token
remains replayable.
* *RFC 9449 (DPoP)* [RFC9449]: Provides application-layer proof-of-
possession using ephemeral, application-managed keys. Applicable
to both public and confidential clients, but binds to the key, not
to the TLS channel, and requires generating and managing a
separate key pair.
* *Token Binding* [RFC8471][RFC8472][RFC8473]: Proposed direct TLS
session binding but required a new TLS extension, was never
specified for Token Exchange, encountered adoption barriers in
browsers and TLS 1.3 transitions, and was ultimately abandoned.
None of these mechanisms provide *TLS-connection-level binding* for
OAuth 2.0 access tokens in mTLS environments. This specification
fills that gap by reusing the existing mTLS key pair (no additional
key generation) and amortizing the proof to once per (token,
connection) pair rather than once per request — delivering stronger
binding than DPoP at lower per-request cost. The binding mechanism
relies on TLS Exporter values, which are available in TLS 1.2, TLS
1.3, and QUIC (via its integrated TLS 1.3 handshake), making the
specification transport-agnostic across modern encrypted transports.
1.2. The Agentic AI Amplifier
The rise of autonomous AI agents dramatically amplifies the bearer
token replay risk. Unlike traditional OAuth flows where a human
initiates discrete requests, agentic AI systems:
* *Autonomously chain API calls*: An agent may invoke hundreds of
API calls across multiple services without human intervention,
each potentially involving RFC 8693 token exchanges.
Krishnan, et al. Expires 24 November 2026 [Page 4]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Perform multi-hop delegation*: Agent A exchanges its token for a
delegated token to call Agent B, which exchanges again for Agent
C. Each hop produces a new bearer token. A single compromise
anywhere in this chain exposes all downstream tokens.
* *Run for extended periods*: Agents operate autonomously for hours
or days, providing a broad window for token interception and
replay.
* *Generate opaque traffic*: Agent-to-agent API calls are fully
automated. A replayed token produces legitimate-looking traffic
that is extremely difficult to distinguish from genuine
requests—there is no human in the loop to detect anomalies.
* *Are susceptible to prompt injection*: A compromised or prompt-
injected LLM agent can exfiltrate bearer tokens via tool calls,
side channels, or log leakage. These tokens are immediately
usable from any connection.
These characteristics make bearer token replay a *first-order threat*
in agentic AI architectures. As autonomous agents increasingly drive
API-to-API traffic volumes that exceed human-initiated requests, per-
request cryptographic overhead becomes a significant scaling concern.
This document addresses this gap by binding access tokens to the mTLS
connection on which they are presented, with proof amortization that
scales to high-volume agent traffic.
1.3. Conventions and Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in BCP
14 [RFC2119] [RFC8174] when, and only when, they appear in all
capitals, as shown here.
This document uses the following terms:
TLS Exporter Value (EKM): A value derived from the TLS handshake
using the mechanism defined in [RFC5705] (for TLS 1.2) or
Section 7.5 of [RFC8446] (for TLS 1.3). EKM stands for Exported
Keying Material. The exporter value is unique to the specific TLS
connection and is available to both endpoints. This document uses
"EKM" and "TLS Exporter value" interchangeably.
Session-Binding Proof: A signed JWT presented alongside the access
token that cryptographically binds the token to the current mTLS
connection via the TLS Exporter value.
Krishnan, et al. Expires 24 November 2026 [Page 5]
Internet-Draft TLS-Session-Bound-Tokens May 2026
Token Exchange: The protocol defined in [RFC8693] for exchanging one
security token for another at an authorization server.
2. TLS Session Binding for Access Tokens
2.1. Overview
This specification defines a proof-of-possession mechanism that binds
OAuth 2.0 access tokens to the mTLS connection on which they are
presented. While applicable to any OAuth 2.0 access token, it is
primarily designed for tokens issued via the Token Exchange protocol
[RFC8693], where multi-hop delegation creates elevated replay risk.
The mechanism operates as follows:
1. The client and resource server establish an mTLS connection.
Both sides derive a TLS Exporter value unique to this connection.
2. When first presenting an access token on a given connection, the
client constructs a *Session-Binding Proof*: a JWT containing the
hash of the access token and the TLS Exporter value.
3. The client signs this JWT with the private key corresponding to
its mTLS client certificate. The same proof MAY be reused for
all subsequent requests using that token on the same connection.
4. The resource server verifies the proof by checking the signature
against the client certificate's public key, confirming the
exporter value matches the current connection, confirming the
token hash matches the presented access token, and confirming the
issuance time is within an acceptable window. The server MUST
perform these checks on every request but MAY use a per-
connection cache to reduce subsequent verifications to a cache
lookup.
A token that requires session binding includes a confirmation method
claim (tls_exp) containing the TLS Exporter label, which signals to
the resource server that the Session-Binding Proof MUST be presented
and verified.
2.2. TLS Exporter Derivation
Both the client and resource server MUST derive the TLS Exporter
value using the following parameters:
* *Label*: EXPORTER-oauth-tls-session-bound
* *Context*: Empty (zero-length)
Krishnan, et al. Expires 24 November 2026 [Page 6]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Length*: 32 octets
For TLS 1.3, the exporter is derived as specified in Section 7.5 of
[RFC8446]. For TLS 1.2, the exporter is derived as specified in
[RFC5705].
Note: By binding to the TLS Exporter rather than the application
traffic keys, the binding remains valid across TLS 1.3 KeyUpdate
operations. Standard key rotation refreshes traffic keys but does
not change the exporter master secret, avoiding unnecessary re-proof
cycles while maintaining strong connection binding.
Note: This mechanism requires both the client and resource server to
access the TLS Exporter derivation function from their TLS library.
Most TLS libraries (OpenSSL, BoringSSL, Go crypto/tls) provide this
API (e.g., SSL_export_keying_material()). However, some application
frameworks and proxy platforms may not yet surface this function to
their HTTP or extension layer. Implementors should verify that their
TLS integration exposes exporter derivation to the layer responsible
for constructing or verifying the Session-Binding Proof.
2.3. Session-Binding Proof Format
The Session-Binding Proof is a JWT with the following structure:
2.3.1. Header
{
"typ": "tls-binding-proof+jwt",
"alg": "ES256",
"x5t#S256": "<base64url SHA-256 thumbprint of client certificate>"
}
The alg value MUST match the key type of the client's mTLS
certificate. The x5t#S256 value MUST be the base64url-encoded
SHA-256 thumbprint of the DER-encoded client certificate, as defined
in [RFC8705].
2.3.2. Payload
The payload contains REQUIRED connection-level claims that prove
session binding, and OPTIONAL per-request claims for intra-session
hardening.
Krishnan, et al. Expires 24 November 2026 [Page 7]
Internet-Draft TLS-Session-Bound-Tokens May 2026
{
"ath": "<base64url SHA-256 hash of the access token>",
"ekm": "<base64url TLS Exporter value>",
"iat": 1710820000,
"jti": "<unique identifier>",
"htm": "POST",
"htu": "/api/resource"
}
Note: The example above shows all defined claims. The ath, ekm, and
iat claims are REQUIRED; the jti, htm, and htu claims are OPTIONAL
per-request claims included only when intra-session hardening is
required.
The payload claims are defined as follows:
2.3.2.1. Connection-Level Claims (REQUIRED)
ath: REQUIRED. The base64url-encoded SHA-256 hash of the ASCII
encoding of the associated access token value.
ekm: REQUIRED. The base64url-encoded TLS Exporter value derived as
specified in Section 2.2.
iat: REQUIRED. The time at which the proof was issued, as a
NumericDate (seconds since the Unix epoch). The resource server
MUST verify that this value is within an acceptable window.
2.3.2.2. Per-Request Claims (OPTIONAL)
The following claims provide intra-session request-level binding.
They are OPTIONAL and intended for deployments that require defense-
in-depth against intra-connection replay. When omitted, the proof
can be reused across all requests on the same connection for the same
token.
jti: OPTIONAL. A unique identifier for the proof. When present,
the resource server MUST verify that this value has not been seen
before within the token's validity period.
htm: OPTIONAL. The HTTP method of the request to which the proof is
attached (e.g., "GET", "POST"). When present, the resource server
MUST verify that it matches the actual request method.
htu: OPTIONAL. The HTTP target URI of the request to which the
proof is attached, without query and fragment parts. When
present, the resource server MUST verify that it matches the
actual request URI.
Krishnan, et al. Expires 24 November 2026 [Page 8]
Internet-Draft TLS-Session-Bound-Tokens May 2026
2.3.3. Signature
The JWT MUST be signed using the private key corresponding to the
client's mTLS certificate. The signature algorithm MUST match the
alg header parameter.
2.4. Token Confirmation Claim
An authorization server that supports TLS-session-bound access tokens
MUST include a cnf (confirmation) claim in the issued access token
(when the token is a JWT) or in the token introspection response.
The cnf claim MUST contain:
{
"cnf": {
"x5t#S256": "<cert-thumbprint>",
"tls_exp": "EXPORTER-oauth-tls-session-bound"
}
}
x5t#S256: REQUIRED. The certificate thumbprint as defined in
[RFC8705].
tls_exp: REQUIRED. A string value containing the TLS Exporter label
that the client and resource server MUST use to derive the
session-binding value. The presence of this claim signals that
the resource server MUST require and verify a Session-Binding
Proof whenever this token is presented. This follows the pattern
of existing cnf members which carry key/binding material rather
than boolean flags (see [RFC7800]).
3. Protocol Flow
3.1. Token Issuance with Session Binding
The mechanism for requesting and issuing TLS-session-bound tokens is
as follows:
1. The client authenticates to the authorization server using mTLS
and requests a token. This MAY be a token exchange per
[RFC8693], a client credentials grant, or any other OAuth 2.0
grant type.
2. The authorization server issues a new access token.
3. If the authorization server policy requires TLS session binding
for this client, it includes the cnf claim with tls_exp set to
the exporter label in the issued token.
Krishnan, et al. Expires 24 November 2026 [Page 9]
Internet-Draft TLS-Session-Bound-Tokens May 2026
4. The client receives the token and notes the tls_exp requirement.
3.2. Authorization Server Behavior
The authorization server determines whether to issue TLS-session-
bound tokens based on per-client configuration. The following client
registration metadata parameter is defined:
tls_session_bound_access_tokens: OPTIONAL. A boolean value
indicating that the authorization server MUST issue TLS-session-
bound access tokens for this client. When set to true, the
authorization server includes the tls_exp confirmation method in
all access tokens issued to this client. Defaults to false.
The authorization server MAY also apply session binding based on:
* Token exchange policy: When the subject_token or actor_token in an
RFC 8693 exchange is itself session-bound, the resulting token
SHOULD also be session-bound to maintain the security property
across the delegation chain.
* Resource server requirements: When a resource server's metadata
indicates that it requires session-bound tokens.
* Risk-based policy: When the requested scope, audience, or
delegation depth exceeds a policy threshold.
3.3. Resource Access with Session-Binding Proof
The following diagram illustrates the complete flow:
Krishnan, et al. Expires 24 November 2026 [Page 10]
Internet-Draft TLS-Session-Bound-Tokens May 2026
Client (with cert C) Resource Server
| |
|=== mTLS handshake (client cert C) =================>|
| |
| [ONCE PER CONNECTION] |
| Both sides derive: |
| EKM = TLS-Exporter( |
| "EXPORTER-oauth-tls-session-bound", |
| "", 32) |
| |
| [ONCE PER (TOKEN, CONNECTION)] |
| Client constructs proof JWT: |
| header = { typ, alg, x5t#S256 } |
| payload = { |
| ath: SHA256(access_token), |
| ekm: EKM, |
| iat: <unix_timestamp> |
| } |
| sig = Sign(C.privateKey, header || payload) |
| |
|--- HTTP Request 1 -------------------------------->|
| Authorization: Bearer <access_token> |
| Session-Binding-Proof: <proof_jwt> |
| |
| Server verifies: |
| 1. sig matches C.publicKey from mTLS |
| 2. ath matches SHA256(access_token) |
| 3. ekm matches server-derived EKM |
| 4. iat within acceptable window |
| (caches: this token is bound to this conn) |
| |
|<-- 200 OK ----------------------------------------|
| |
|--- HTTP Request 2 (same token, same proof) ------->|
| Authorization: Bearer <access_token> |
| Session-Binding-Proof: <proof_jwt> (reused) |
| |
| Server verifies (per request): |
| - looks up (conn_id, ath) in cache |
| - cache hit: binding verified for THIS conn |
| |
|<-- 200 OK ----------------------------------------|
Krishnan, et al. Expires 24 November 2026 [Page 11]
Internet-Draft TLS-Session-Bound-Tokens May 2026
3.3.1. Performance Characteristics
Because the Session-Binding Proof contains only connection-level
claims (ekm, ath, iat), the client constructs and signs the proof
*once per (token, connection) pair* and reuses it for all subsequent
requests on that connection. The iat value reflects the time of
proof construction, not the time of any particular request. This
provides a significant advantage over DPoP:
* *Proof construction (client-side)*: One JWT signature per
connection per token. DPoP requires one JWT signature per
request.
* *Proof verification (server-side)*: Full JWT verification on first
presentation; the server MAY cache the verified (connection, ath)
binding and reduce subsequent verifications to a cache lookup.
DPoP requires full JWT verification per request.
* *TLS Exporter derivation*: Once per connection, cached by both
sides.
* *No separate key management*: Unlike DPoP, which requires
generating, storing, and rotating an ephemeral key pair
independent of TLS, this specification reuses the mTLS key pair.
This eliminates an entire key lifecycle from the implementation.
When OPTIONAL per-request claims (jti, htm, htu) are included, the
proof must be constructed per request (similar to DPoP). Deployments
choose between per-connection efficiency and per-request intra-
session hardening based on their threat model.
This amortization benefit assumes that mTLS connections are *long-
lived* relative to the number of requests. If a deployment creates a
new mTLS connection for every HTTP request-response cycle, the proof
must be constructed on every request and the per-request cost becomes
equivalent to DPoP. Deployments SHOULD use persistent connections
(e.g., HTTP/2 or HTTP/1.1 with keep-alive) so that a single mTLS
connection carries many requests. This is already standard practice
in service mesh and workload-to-workload environments where mTLS is
terminated at a sidecar or gateway.
3.3.2. Mid-Session Token Binding
A client MAY bind multiple tokens to a single mTLS connection
concurrently. Each token requires its own Session-Binding Proof with
a distinct ath (token hash) and the same ekm (TLS Exporter value)
from the current connection.
Krishnan, et al. Expires 24 November 2026 [Page 12]
Internet-Draft TLS-Session-Bound-Tokens May 2026
This is common in workload-to-workload scenarios: a single mTLS
connection between two microservices may carry requests on behalf of
many different users, each with a distinct access token obtained via
separate token exchanges. For example, 100 users interacting with an
AI agentic service that delegates to a downstream service — each
user's token is bound to the same mTLS connection via a separate
proof. Similarly, when a token is refreshed or rotated mid-session,
the client constructs a fresh proof for the new token without
requiring a new mTLS connection.
The resource server MAY cache verified (connection_id, ath) bindings
to avoid re-verifying proofs on subsequent requests. Each cache
entry is small (a connection identifier plus a 32-byte hash), so even
connections carrying hundreds of tokens have modest memory
requirements. Caching is an optimization, not a requirement: if the
server does not cache (or evicts entries under memory pressure), it
simply falls back to full JWT signature verification on every request
— equivalent to the per-request cost of DPoP.
When the client presents the access token to a resource server:
1. The client establishes an mTLS connection with the resource
server.
2. The client derives the TLS Exporter value for the current
connection (computed once and cached).
3. The client constructs a Session-Binding Proof JWT as specified in
Section 2.3. If the proof does not contain per-request claims,
the client MAY reuse the same proof for all subsequent requests
using this token on this connection.
4. The client sends the HTTP request with:
* The access token in the Authorization header: Authorization:
Bearer <access_token>
* The Session-Binding Proof in the Session-Binding-Proof header:
Session-Binding-Proof: <proof_jwt>
5. The resource server performs the following verifications:
a. *Standard mTLS verification*: Verifies the client certificate
as part of the TLS handshake. b. *Token validation*: Validates
the access token (signature, expiration, audience, etc.). c.
*Certificate binding*: Verifies that the x5t#S256 in the token's
cnf claim matches the presented client certificate. d. *TLS
binding required*: Checks that cnf.tls_exp is present and a
Krishnan, et al. Expires 24 November 2026 [Page 13]
Internet-Draft TLS-Session-Bound-Tokens May 2026
Session-Binding Proof is present. e. *Proof signature*: Verifies
the proof JWT signature against the public key in the client
certificate. f. *Exporter match*: Compares the ekm claim in the
proof against the TLS Exporter value for the current connection
and confirms they match. g. *Token hash*: Computes SHA-256 of
the presented access token and confirms it matches the ath claim.
h. *Timestamp*: Confirms iat is within an acceptable window. i.
*Per-request claims* (when present): Confirms htm and htu match
the actual request, and jti has not been seen before.
The resource server MUST ensure that steps (a) through (i) are
satisfied for every request. This is critical: an attacker who
obtains a stolen token and proof may attempt to inject them on
their own mTLS connection to the same resource server.
Verification ensures the proof signature is checked against the
certificate from _this_ connection's handshake, which will not
match the attacker's certificate.
To satisfy this requirement efficiently, the resource server MAY
cache verified bindings keyed on (connection_id, ath). On
subsequent requests with the same token on the same connection, a
cache hit confirms the binding was already verified; a cache miss
(different connection, different token, or first presentation)
triggers full verification. This cache is inherently per-
connection, so a stolen proof presented on a different connection
will always miss and fail full verification. When a TLS
connection terminates, the resource server SHOULD purge all cache
entries associated with that connection, as the bindings are no
longer valid.
6. If all verifications succeed, the resource server processes the
request. If any verification fails, the resource server MUST
reject the request as specified in Section 3.5.
3.4. Token Introspection Considerations
When token introspection [RFC7662] is used, the introspection
response MUST include the cnf claim with the tls_exp field. This
allows resource servers that do not have direct access to the token's
claims (e.g., opaque tokens) to determine whether session binding is
required.
3.5. Error Responses
When verification of the Session-Binding Proof fails or the proof is
missing, the resource server MUST respond with HTTP 401 and include a
WWW-Authenticate header with the appropriate error code:
Krishnan, et al. Expires 24 November 2026 [Page 14]
Internet-Draft TLS-Session-Bound-Tokens May 2026
WWW-Authenticate: Bearer error="invalid_proof",
error_description="Session-Binding Proof verification failed:
exporter mismatch"
WWW-Authenticate: Bearer error="use_session_binding",
error_description="This token requires a Session-Binding Proof"
The following error code values are defined:
invalid_proof: The Session-Binding Proof is malformed or failed
verification. This includes signature verification failure,
exporter mismatch, stale iat, jti reuse, ath mismatch, and htm/htu
mismatch.
use_session_binding: The access token requires a Session-Binding
Proof (the cnf.tls_exp claim is present) but no Session-Binding-
Proof header was provided. This error signals to the client that
it must construct and present a proof.
The resource server SHOULD include an error_description parameter
with a human-readable explanation of the specific verification
failure.
4. Integration with Existing Mechanisms
4.1. Relationship to RFC 8705 (mTLS-Bound Tokens)
This specification extends RFC 8705 by adding session-level binding
on top of certificate binding. The x5t#S256 claim from RFC 8705 is
reused. Deployments MAY support both mechanisms simultaneously: RFC
8705 provides certificate binding, while this specification adds per-
connection session binding.
4.2. Relationship to RFC 9449 (DPoP)
DPoP and this specification address similar goals (proof-of-
possession) but differ in binding targets, key management, per-
request cost, and the provenance of the binding key:
* *Binding target*: DPoP binds tokens to an ephemeral application-
layer key. This specification binds tokens to both the client
identity (X.509 certificate) and the specific TLS connection
(Exporter value). An attacker who exfiltrates a DPoP-bound token
and the associated DPoP key can replay the token from any network
location; an attacker who exfiltrates a session-bound token
cannot, because reproducing the TLS Exporter value requires being
on the same TLS connection.
Krishnan, et al. Expires 24 November 2026 [Page 15]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Key management*: DPoP requires generating, storing, and rotating
an ephemeral key pair independent of TLS. This specification
reuses the mTLS key pair already established during the handshake,
eliminating the entire DPoP key lifecycle.
* *Per-request overhead*: DPoP requires constructing and signing a
new proof JWT for every HTTP request. This specification
constructs the proof *once per (token, connection)* and reuses it
across all requests on that connection (when per-request claims
are omitted). This is a fundamental efficiency advantage in high-
throughput workload-to-workload scenarios.
* *Identity provenance*: A DPoP key is self-generated by the client
with no provenance beyond the client's own assertion — any process
can create one. This specification reuses the mTLS key, which in
workload deployments is issued through a verifiable attestation
chain. When the mTLS credential is a standards-based workload
identity (e.g., a SPIFFE X.509-SVID issued by SPIRE), the key's
issuance was conditioned on node attestation followed by workload
attestation. The Session-Binding Proof is therefore not merely
"possession of a key" but "possession of a key whose issuance was
verified against the workload's identity and execution context."
See Appendix C for details.
* *Applicability*: DPoP is particularly valuable for *public
clients* that cannot use mTLS. This specification is designed for
*confidential clients and workloads* that already use mTLS. The
two mechanisms are complementary rather than competing.
In summary, for mTLS-capable environments, this specification
provides stronger security (connection-level binding vs. key-level
binding), lower per-request cost (amortized proof vs. per-request
proof), simpler implementation (no separate key management), and —
when backed by a workload identity system — proof whose binding key
has verifiable provenance through an attestation chain.
4.3. Relationship to Token Binding (RFC 8471-8473)
Token Binding [RFC8471][RFC8472][RFC8473] proposed direct TLS session
binding for OAuth and HTTP but did not gain traction in industry.
This specification addresses the same core problem — preventing
cross-connection token replay — while avoiding the specific
deployment barriers that Token Binding encountered.
* *No new TLS extension.* Token Binding required a dedicated TLS
negotiation extension ([RFC8472]) that had to be implemented and
enabled by browser vendors and TLS stacks. This specification
uses the TLS Exporter mechanism ([RFC5705], [RFC8446]
Krishnan, et al. Expires 24 November 2026 [Page 16]
Internet-Draft TLS-Session-Bound-Tokens May 2026
Section 7.5), which is already available in every compliant TLS
1.2, TLS 1.3, and QUIC implementation without any extension or
negotiation.
* *Session-scoped binding, not persistent client keys.* Token
Binding associated tokens with long-lived, client-managed signing
keys, creating persistent client identifiers across connections.
This specification binds tokens to the TLS Exporter value, which
is unique to each connection and exists only for the lifetime of
that session. No persistent identifier is created beyond the TLS
session itself.
* *Specified for Token Exchange.* Token Binding was not defined for
use with RFC 8693 Token Exchange delegation chains. This
specification is primarily motivated by exactly that scenario:
each hop in a multi-hop agentic delegation chain produces a new
bearer token, and session binding contains the blast radius of any
single compromise to the specific connection on which that token
is presented.
* *Explicit deployment scope constraint.* Token Binding encountered
complexity in proxy and intermediary topologies, leading to
additional specifications ([RFC8473]) to handle referred binding
over HTTP. This specification avoids that complexity by
explicitly restricting the binding guarantee to deployments where
the verifier is a direct endpoint of the client TLS connection —
either the resource server itself or a co-located sidecar.
Forwarding the EKM through a remote intermediary would change the
security property from session binding to proxy-attested binding,
which is a categorically weaker guarantee. See Section 5 for
details.
This specification is not applicable to browser-based web clients, as
browsers do not currently expose TLS Exporter values to JavaScript.
It is targeted at server-to-server, service-mesh, and agentic AI
workloads where both endpoints have direct access to the TLS stack.
4.4. Relationship to WIMSE WIT/WPT
The WIMSE Workload Identity Token (WIT) and Workload Proof Token
(WPT) defined in [I-D.ietf-wimse-s2s-protocol] provide a proof-of-
possession mechanism for workload-to-workload communication. This
specification is compatible with WIMSE and addresses a complementary
problem.
WIT/WPT and this specification operate at different layers and answer
different questions:
Krishnan, et al. Expires 24 November 2026 [Page 17]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *WIMSE WIT/WPT*: "Is this workload who it claims to be, and does
it currently hold the private key corresponding to its identity?"
(application-layer identity assertion and per-request proof of
possession)
* *This specification*: "Is this OAuth access token being presented
on the TLS connection that this workload authenticated?" (TLS-
channel-level token binding scoped to a specific connection)
Neither mechanism alone provides both properties. In a workload-to-
workload flow where both are deployed, WIT/WPT establishes and proves
the workload's identity at the application layer while this
specification ensures the OAuth authorization token is
cryptographically bound to the specific connection that workload
established. Together they close the complete chain: from attested
workload identity to authorized, connection-scoped API access.
WPT is a per-request application-layer proof; this specification
amortizes the proof to once per (token, connection) pair at the TLS-
channel layer. The two mechanisms are not competing alternatives —
deployments that require both workload identity proof and connection-
level token binding SHOULD use both.
4.5. Relationship to Transitive Attestation
The Transitive Attestation profile
[I-D.draft-mw-wimse-transitive-attestation] addresses a complementary
problem: binding an identity to a verified execution environment
("Proof of Residency"). While this specification binds tokens to a
TLS connection to prevent network-level replay, Transitive
Attestation binds identities to a hardware-rooted host to prevent
credential export. In high-assurance deployments, both mechanisms
MAY be combined: Transitive Attestation ensures the token is used
from the correct host, and TLS session binding ensures it is used on
the correct connection.
4.6. Relationship to RFC 8693 (Token Exchange)
This specification does not modify the token exchange protocol
itself. The authorization server's token exchange endpoint continues
to operate as specified in [RFC8693]. The session binding is applied
to the _resulting_ access token through the cnf claim. While this
specification is applicable to any OAuth 2.0 access token, RFC 8693
Token Exchange is a primary motivator: each hop in a delegation chain
produces a new bearer token, and session binding contains the blast
radius of any single token compromise to the specific TLS connection
on which it is presented.
Krishnan, et al. Expires 24 November 2026 [Page 18]
Internet-Draft TLS-Session-Bound-Tokens May 2026
5. Deployment Scope
5.1. Co-Located TLS Termination (Supported)
This specification requires that the entity verifying the Session-
Binding Proof be an endpoint of the TLS connection on which the token
is presented, or a sidecar co-located at the same trust boundary as
the resource server that terminates that connection directly. In
both cases, the verifier derives the EKM from the TLS session it
directly participates in, and no EKM forwarding is required.
This covers two deployment models:
* *Pass-through mTLS*: TLS is terminated at the application server
itself. Both the client and resource server derive the EKM
directly from the same connection.
* *Co-located sidecar*: TLS is terminated at a sidecar (e.g., Envoy)
running in the same pod or VM as the resource server. The sidecar
derives the EKM directly and verifies the proof before passing the
request to the application. This is the recommended deployment
model for agentic AI environments. See Appendix B.
5.2. Remote TLS Termination (Out of Scope)
Deployments where TLS is terminated at a remote intermediary — such
as a standalone load balancer or API gateway that is not co-located
with the resource server — are outside the scope of this
specification.
In such topologies, the EKM cannot be conveyed to the verifier
without introducing a trust dependency on the intermediary's
assertion of what the EKM was. This changes the security property
from "bound to the TLS session" to "bound to what a proxy claims the
TLS session's EKM was" — a weaker and categorically different
guarantee that undermines the core security claim of this
specification. Allowing EKM to escape the session boundary via an
HTTP header would recreate the kind of mediated-trust complexity that
contributed to TLS Token Binding's adoption failures
[RFC8471][RFC8472][RFC8473].
Deployments requiring TLS-terminating intermediaries SHOULD place the
TLS termination point and the Session-Binding Proof verifier within
the same trust boundary — for example, by running an Envoy-based
sidecar as both the TLS endpoint and the proof verifier, co-located
with the resource server backend.
Krishnan, et al. Expires 24 November 2026 [Page 19]
Internet-Draft TLS-Session-Bound-Tokens May 2026
5.3. QUIC and HTTP/3
This specification is directly applicable to QUIC ([RFC9000]) and
HTTP/3 transports. QUIC integrates TLS 1.3 into its handshake
([RFC9001]), and TLS Exporter values are available for use with QUIC
connections as specified in Section 4.2 of [RFC9001].
The following considerations apply when using this specification over
QUIC:
* *Exporter derivation*: The TLS Exporter MUST be derived using the
1-RTT exporter keys (post-handshake). Proofs MUST NOT be
constructed using 0-RTT exporter values, as these are derived from
pre-shared keys and provide weaker binding guarantees.
* *Connection migration*: QUIC connections may migrate across
network paths (different IP addresses and ports) without a new TLS
handshake. The TLS state — and therefore the EKM — persists
across migrations. Session-Binding Proofs remain valid after
connection migration.
* *Client authentication*: In TLS 1.3 over QUIC, post-handshake
client authentication is not supported (Section 4.4 of [RFC9001]).
Client certificates MUST be presented during the initial
handshake. This simplifies the binding model: the client identity
is fixed at connection establishment.
* *Long-lived connections*: QUIC connections are designed to be
long-lived and resilient to network changes, aligning naturally
with the long-lived connection guidance in Section 3.3.1.
6. Security Considerations
This section addresses security considerations in addition to those
described in the OAuth 2.0 Security Best Current Practice
[I-D.ietf-oauth-security-topics].
6.1. Threat Model
The following table summarizes the threats addressed by this
specification, the specific mechanism that mitigates each threat, and
whether DPoP provides equivalent protection.
6.1.1. T1: Cross-Connection Token Replay
* *Threat*: An attacker intercepts a bearer token and presents it on
a different TLS connection.
Krishnan, et al. Expires 24 November 2026 [Page 20]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Mitigation*: The ekm claim in the proof is derived from the TLS
handshake transcript and is unique per connection. The resource
server compares it against its own locally-derived EKM. A
replayed proof will fail the EKM comparison.
* *DPoP equivalent*: No. DPoP binds to an application-layer key,
not to the TLS connection. If the DPoP key is also exfiltrated,
replay succeeds.
6.1.2. T2: Cross-Host Token Replay
* *Threat*: An attacker on a different host obtains a bearer token
and attempts to use it.
* *Mitigation*: The proof is signed with the client's mTLS private
key. The resource server verifies the signature against the
public key from the current mTLS handshake. A different host
presents a different certificate, so signature verification fails.
* *DPoP equivalent*: Partial. DPoP binds to an ephemeral key, which
may not be hardware-protected.
6.1.3. T3: Token Exfiltration via LLM Prompt Injection
* *Threat*: A compromised or prompt-injected AI agent exfiltrates a
bearer token via tool calls, side channels, or log leakage.
* *Mitigation*: The attacker obtains the token but cannot produce a
valid proof. When deployed with a security sidecar (see
Appendix B), the mTLS private key resides in a separate process
inaccessible to the agent's LLM runtime. When deployed without a
sidecar (direct integration, see Appendix C), the private key
resides in the same process as the agent; key isolation in this
variant depends on platform-level controls such as hardware-backed
key storage. In both cases, even if the attacker also obtains the
proof, the EKM will not match on a different connection.
* *DPoP equivalent*: No. DPoP keys are application-layer and
typically reside in the same process as the agent, making co-
exfiltration with the token likely.
6.1.4. T4: Token + Proof Exfiltration (Both Stolen)
* *Threat*: An attacker obtains both the access token and the
Session-Binding Proof, and attempts to use them on their own mTLS
connection to the same resource server.
Krishnan, et al. Expires 24 November 2026 [Page 21]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Mitigation*: Two independent protections apply: (1) the
attacker's mTLS certificate differs from the original client's, so
the proof signature does not match the public key from the
attacker's handshake — the resource server verifies the proof
against the certificate presented on _this_ connection; (2) even
if the attacker somehow presents the same certificate, a different
TLS connection produces a different EKM, so the ekm claim does not
match the server's locally-derived value.
* *DPoP equivalent*: No. If both the DPoP proof and the DPoP key
are exfiltrated, the attacker can replay from any connection.
6.1.5. T5: Multi-Hop Delegation Chain Compromise
* *Threat*: A token is stolen at one hop in a delegation chain
(A→B→C→D) and replayed at a different hop.
* *Mitigation*: Each hop uses a distinct mTLS connection with a
distinct EKM. A token bound to one hop's connection cannot be
replayed on another.
* *DPoP equivalent*: Partial. DPoP binds per key, but the key is
not inherently tied to a specific hop's connection.
6.2. Residual Risks
6.2.1. Intra-Connection Replay
Within the same TLS connection, an attacker with access to the
channel (e.g., a compromised middleware component) could observe and
replay requests. This risk is mitigated by including the OPTIONAL
per-request claims:
* The jti claim, which provides per-proof uniqueness when the server
maintains a replay cache.
* The htm and htu claims, which bind the proof to a specific HTTP
method and URI.
* Short iat validity windows that limit the temporal scope of any
replay.
Deployments that require intra-connection replay protection SHOULD
include per-request claims.
Krishnan, et al. Expires 24 November 2026 [Page 22]
Internet-Draft TLS-Session-Bound-Tokens May 2026
6.2.2. Compromised Private Key
If the client's mTLS private key is compromised, the attacker can
produce valid proofs. Mitigations include:
* *Hardware-backed key storage*: Storing the private key in a TPM,
HSM, or TEE prevents software-level exfiltration.
* *Short-lived certificates*: Using short-lived credentials (e.g.,
SPIFFE SVIDs with hourly or shorter expiry) limits the window
during which a compromised key can be exploited.
* *Transitive Attestation*:
[I-D.draft-mw-wimse-transitive-attestation] binds identity to a
verified execution context, providing evidence if the execution
environment is tampered with.
7. IANA Considerations
7.1. OAuth Token Confirmation Methods
This specification registers the following confirmation method in the
IANA "OAuth Token Confirmation Methods" registry established by
[RFC7800]:
* *Confirmation Method Value*: tls_exp
* *Confirmation Method Description*: TLS Exporter Session Binding
* *Change Controller*: IETF
* *Reference*: [this document]
7.2. OAuth Dynamic Client Registration Metadata
This specification registers the following client metadata value:
* *Client Metadata Name*: tls_session_bound_access_tokens
* *Client Metadata Description*: Boolean indicating the client
requires TLS-session-bound access tokens
* *Change Controller*: IETF
* *Reference*: [this document]
Krishnan, et al. Expires 24 November 2026 [Page 23]
Internet-Draft TLS-Session-Bound-Tokens May 2026
7.3. HTTP Header Fields
This specification registers the following HTTP header fields:
7.3.1. Session-Binding-Proof
* *Header Field Name*: Session-Binding-Proof
* *Status*: permanent
* *Reference*: [this document]
7.4. TLS Exporter Label
This specification registers the following TLS Exporter label in the
IANA "TLS Exporter Labels" registry:
* *Value*: EXPORTER-oauth-tls-session-bound
* *DTLS-OK*: N
* *Recommended*: Y
* *Reference*: [this document]
8. Normative References
[RFC5705] Rescorla, E., "Keying Material Exporters for Transport
Layer Security (TLS)", March 2010,
<https://www.rfc-editor.org/rfc/rfc5705>.
[RFC7662] Richer, J., "OAuth 2.0 Token Introspection", October 2015,
<https://www.rfc-editor.org/rfc/rfc7662>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", May 2017,
<https://www.rfc-editor.org/rfc/rfc8174>.
[RFC9449] Fett, D., Campbell, B., Bradley, J., Lodderstedt, T.,
Jones, M., and D. Waite, "OAuth 2.0 Demonstrating Proof of
Possession (DPoP)", September 2023,
<https://www.rfc-editor.org/rfc/rfc9449>.
[RFC8446] Rescorla, E., "The Transport Layer Security (TLS) Protocol
Version 1.3", August 2018,
<https://www.rfc-editor.org/rfc/rfc8446>.
Krishnan, et al. Expires 24 November 2026 [Page 24]
Internet-Draft TLS-Session-Bound-Tokens May 2026
[RFC7800] Jones, M., Bradley, J., and H. Tschofenig, "Proof-of-
Possession Key Semantics for JSON Web Tokens (JWTs)",
April 2016, <https://www.rfc-editor.org/rfc/rfc7800>.
[RFC8693] Jones, M., Nadalin, A., Campbell, B., Bradley, J., and C.
Mortimore, "OAuth 2.0 Token Exchange", January 2020,
<https://www.rfc-editor.org/rfc/rfc8693>.
[RFC8705] Campbell, B., Bradley, J., Sakimura, N., and T.
Lodderstedt, "OAuth 2.0 Mutual-TLS Client Authentication
and Certificate-Bound Access Tokens", February 2020,
<https://www.rfc-editor.org/rfc/rfc8705>.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", March 1997,
<https://www.rfc-editor.org/rfc/rfc2119>.
9. Informative References
[I-D.ietf-wimse-s2s-protocol]
Howard, P., "WIMSE Service to Service Authentication", 21
October 2024, <https://datatracker.ietf.org/doc/html/
draft-ietf-wimse-s2s-protocol>.
[RFC9000] Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed
and Secure Transport", May 2021,
<https://www.rfc-editor.org/rfc/rfc9000>.
[I-D.ietf-oauth-security-topics]
Lodderstedt, T., Bradley, J., Labunets, A., and D. Fett,
"OAuth 2.0 Security Best Current Practice", April 2025,
<https://datatracker.ietf.org/doc/html/draft-ietf-oauth-
security-topics>.
[RFC8471] Popov, A., Nystrom, M., Balfanz, D., and J. Hodges, "The
Token Binding Protocol Version 1.0", October 2018,
<https://www.rfc-editor.org/rfc/rfc8471>.
[RFC8472] Popov, A., Nystrom, M., and D. Balfanz, "Transport Layer
Security (TLS) Extension for Token Binding Protocol
Negotiation", October 2018,
<https://www.rfc-editor.org/rfc/rfc8472>.
[RFC8473] Popov, A., Nystrom, M., Balfanz, D., Harper, N., and J.
Hodges, "Token Binding over HTTP", October 2018,
<https://www.rfc-editor.org/rfc/rfc8473>.
Krishnan, et al. Expires 24 November 2026 [Page 25]
Internet-Draft TLS-Session-Bound-Tokens May 2026
[RFC9001] Thomson, M. and S. Turner, "Using TLS to Secure QUIC", May
2021, <https://www.rfc-editor.org/rfc/rfc9001>.
[I-D.draft-mw-wimse-transitive-attestation]
Krishnan, R., "Transitive Attestation for Sovereign
Workloads: A WIMSE Profile", March 2026,
<https://datatracker.ietf.org/doc/html/draft-mw-wimse-
transitive-attestation>.
Appendix A. Contributors
The following individuals have contributed to this document:
Roman Zabicki
JPMorgan Chase & Co
Email: roman.zabicki@chase.com
Jack McCallie
JPMorgan Chase & Co
Email: jack.mccallie@chase.com
Jonian Musa JPMorgan Chase & Co
Email: jonian.musa@chase.com
Appendix B. Sidecar Deployment for Agentic AI
In agentic AI architectures, a common deployment pattern places a
security sidecar (e.g., Envoy, Istio proxy, or a purpose-built agent
gateway) alongside each AI agent workload. This appendix describes
how TLS-session-bound tokens integrate with this pattern and the
resulting security benefits.
B.1. Architecture
The following diagram shows the deployment layout. The AI agent and
security sidecar run in the same pod or VM. The sidecar holds the
mTLS private key, manages all authenticated outbound connections, and
maintains a cache of signed proofs. A single mTLS connection to a
remote resource server may carry requests on behalf of many different
users, each with a distinct access token; the sidecar constructs and
signs a proof once per (token, connection) pair and reuses it for
subsequent requests with the same token.
Krishnan, et al. Expires 24 November 2026 [Page 26]
Internet-Draft TLS-Session-Bound-Tokens May 2026
+----------------------------------------------------+
| Pod / VM |
| |
| +------------------+ +-------------------+ |
| | AI Agent | | Security | |
| | (app logic, | | Sidecar | |
| | LLM runtime) | | (Envoy etc.) | |
| | | | | |
| | Holds: | | Holds: | |
| | - access tokens | | - mTLS private key| |
| | (per user) | | - X.509 cert | |
| | - prompts | | - EKM cache | |
| | - app context | | - proof cache | |
| | | | (per token) | |
| | Does NOT hold: | | Constructs: | |
| | - private keys | | - Binding Proof | |
| +--------+---------+ +--------+----------+ |
| | plaintext HTTP | |
| | (loopback/UDS) | mTLS |
| +---------->-------------+| |
| | |
+-------------------------------------+---------------+
|
v
Remote Resource Server
B.2. Request Flow
The following diagram shows how requests flow from the AI agent
through the sidecar to the remote resource server. The sidecar
transparently adds the Session-Binding Proof and reuses it for
subsequent requests with the same token.
AI Agent Security Sidecar Resource Server
| | |
| |=== mTLS handshake ======>|
| | |
| | [ONCE PER CONNECTION] |
| | Both sides derive: |
| | EKM = TLS-Exporter( |
| | "EXPORTER-oauth- |
| | tls-session- |
| | bound", "", 32) |
| | |
| (1) HTTP Request 1 | |
| Authorization: | |
| Bearer <token> | |
| (plaintext, local) | |
Krishnan, et al. Expires 24 November 2026 [Page 27]
Internet-Draft TLS-Session-Bound-Tokens May 2026
+---------------------->| |
| | |
| (2) Sidecar (ONCE PER TOKEN, CONN):|
| - Computes ath = SHA256(token) |
| - Constructs proof JWT: |
| { ath, ekm: EKM, iat } |
| - Signs with mTLS private key |
| | |
| | (3) mTLS Request |
| | Authorization: |
| | Bearer <token> |
| | Session-Binding-Proof: |
| | <proof_jwt> |
| +------------------------>|
| | |
| | (4) Server verifies: |
| | 1. sig matches cert |
| | 2. ath matches token |
| | 3. ekm matches EKM |
| | 4. iat in window |
| | (caches binding) |
| | |
| | (5) 200 OK |
| |<------------------------+
| (6) 200 OK | |
|<----------------------+ |
| | |
| (7) HTTP Request 2 | |
| Authorization: | |
| Bearer <token> | |
+---------------------->| |
| | |
| (8) Sidecar reuses proof |
| | |
| | (9) mTLS Request |
| | Authorization: |
| | Bearer <token> |
| | Session-Binding-Proof: |
| | <proof_jwt> (reused) |
| +------------------------>|
| | |
| | (10) Cache hit: |
| | binding verified |
| | |
| | (11) 200 OK |
| |<------------------------+
| (12) 200 OK | |
|<----------------------+ |
Krishnan, et al. Expires 24 November 2026 [Page 28]
Internet-Draft TLS-Session-Bound-Tokens May 2026
| | |
B.3. HTTP/2 Multi-User Multiplexing
In a typical agentic deployment, Agent A serves multiple users
concurrently. Each user's request triggers an on-behalf-of (OBO)
token exchange, producing a distinct access token. All outbound
requests to Agent B share a single HTTP/2 mTLS connection,
multiplexed across streams:
Agent A Sidecar Agent B
| | |
| |=== mTLS (one conn) ===>|
| | EKM derived once |
| | |
User 1: | |
|--Token_1-->| |
| |--(stream 1)----------->|
| | Bearer: Token_1 |
| | Proof: {ath_1,ekm} |
| | (signed once) |
| | |
User 2: | |
|--Token_2-->| |
| |--(stream 3)----------->|
| | Bearer: Token_2 |
| | Proof: {ath_2,ekm} |
| | (signed once) |
| | |
User 1 again: | |
|--Token_1-->| |
| |--(stream 5)----------->|
| | Bearer: Token_1 |
| | Proof: reused |
| | |
Each token is signed *once* when first seen on this connection.
Subsequent requests with the same token reuse the cached proof. With
N users and M requests per user, the total signing cost is N (one per
token) rather than N×M (one per request as with DPoP).
B.4. Security Benefits
This architecture provides defense-in-depth against agentic AI threat
vectors:
Krishnan, et al. Expires 24 November 2026 [Page 29]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Prompt injection token exfiltration*: Even if a compromised LLM
exfiltrates an access token via tool calls, log leakage, or side
channels, the attacker cannot produce a valid Session-Binding
Proof. The mTLS private key resides exclusively in the sidecar
process, which is not accessible to the agent's application logic
or LLM runtime.
* *Key isolation*: The agent never has access to the signing key.
The sidecar can enforce hardware-backed key storage (TPM, HSM)
independently of the agent's runtime environment.
* *Transparent integration*: The agent application code requires no
modifications beyond standard OAuth token handling. The session
binding is entirely transparent—the sidecar intercepts outbound
requests and adds the proof.
* *Centralized policy enforcement*: The sidecar can apply additional
policy checks (token scoping, rate limiting, destination
allowlisting) before constructing the proof, providing a security
control plane for agent traffic.
* *Audit boundary*: All authenticated outbound traffic passes
through the sidecar, providing a natural audit point for logging
which tokens were used, to which destinations, and when.
B.5. Relationship to Existing Infrastructure
This deployment model aligns with the service mesh architecture used
in SPIFFE/SPIRE environments, where the sidecar already manages
workload identity certificates. When combined with Transitive
Attestation [I-D.draft-mw-wimse-transitive-attestation], the sidecar
can additionally attest that the agent is running in a verified
execution environment while simultaneously binding all tokens to the
active TLS connection.
B.6. Implementation Considerations
The sidecar must access the TLS Exporter value from the mTLS
connection it terminates in order to construct the Session-Binding
Proof. See Section 2.2 for the general implementation note on TLS
Exporter API availability across TLS libraries and frameworks.
B.6.1. Connection Lifecycle Management
Sidecars that implement this specification should handle TLS
connection lifecycle events to manage the EKM and proof caches
correctly. A typical implementation uses a layered architecture:
Krishnan, et al. Expires 24 November 2026 [Page 30]
Internet-Draft TLS-Session-Bound-Tokens May 2026
* *Connection-level layer* (e.g., a network filter or transport
callback): Derives the EKM when a new mTLS connection is
established and stores it in connection-scoped state. Registers a
callback for connection close events.
* *Request-level layer* (e.g., an HTTP filter): Reads the cached EKM
from the connection-scoped state, constructs or retrieves the
cached proof for each token, and injects the Session-Binding-Proof
header into outbound requests.
When a TLS connection terminates, all connection-scoped state —
including the EKM and all cached proofs for that connection — SHOULD
be purged automatically. Frameworks that support connection-scoped
storage (e.g., per-connection filter state) provide this cleanup
naturally without requiring explicit invalidation logic.
When a connection to the upstream resource server is lost and a new
one is established, the sidecar derives a fresh EKM from the new
handshake and constructs new proofs. Previously cached proofs are
invalid because they contain the old connection's EKM.
Appendix C. SPIFFE/SPIRE Integration
This appendix is non-normative. It describes how this specification
integrates with SPIFFE/SPIRE workload identity infrastructure and the
additional security properties that result when the mTLS certificate
is a SPIFFE X.509-SVID.
C.1. Identity-Provenance Property
When the mTLS certificate is an arbitrary X.509 certificate, this
specification provides connection-scoped token binding: the token
cannot be replayed on a different TLS connection. When the mTLS
certificate is a SPIFFE X.509-SVID issued by a SPIRE deployment, the
binding acquires a richer semantic property.
A SPIFFE X.509-SVID is issued only after a two-stage attestation
process:
1. *Node attestation*: The SPIRE agent running on the node proves to
the SPIRE server that the node is what it claims to be (e.g., via
cloud provider instance identity documents, TPM attestation, or
Kubernetes service account tokens).
2. *Workload attestation*: The SPIRE agent verifies the properties
of the workload process requesting the SVID (e.g., binary hash,
Unix UID/GID, Kubernetes pod labels, container image digest)
against policies registered in the SPIRE server.
Krishnan, et al. Expires 24 November 2026 [Page 31]
Internet-Draft TLS-Session-Bound-Tokens May 2026
The resulting SVID is short-lived (typically between one minute and
one hour, depending on deployment policy) and is automatically
rotated by the SPIRE agent before expiry.
When a Session-Binding Proof is signed by the private key of a SPIFFE
X.509-SVID, the resource server's verification implies the following
chain:
* The access token was presented on a specific TLS connection (EKM
binding).
* That connection was established by the workload whose private key
signed the proof.
* That workload was attested by a SPIRE agent as running on an
attested node.
* The SVID was valid and within its short-lived issuance window at
the time of proof construction.
This property — authorization bound to attested workload identity —
is not achievable with DPoP, whose key has no provenance beyond self-
generation, or with RFC 8705 alone, which binds to a certificate
thumbprint but does not constrain how the certificate was issued or
how long it has been valid. The combination of this specification
with SPIFFE/SPIRE delivers identity-rooted proof of possession: the
binding key is not "a key the client generated" but "a key issued to
a specific attested workload through a verifiable trust chain."
C.2. Deployment Variants
C.2.1. Without Sidecar (Direct Integration)
In this variant, the workload process itself obtains the SVID via the
SPIFFE Workload API and uses it directly for mTLS connections. The
workload also constructs the Session-Binding Proof using the SVID
private key.
Krishnan, et al. Expires 24 November 2026 [Page 32]
Internet-Draft TLS-Session-Bound-Tokens May 2026
SPIRE Server
|
| (node + workload attestation)
|
SPIRE Agent (node-level)
|
| Workload API (UDS)
|
+--------------------------------------------------+
| Pod / VM |
| |
| +--------------------------------------------+ |
| | Workload (AI Agent / Service) | |
| | | |
| | Holds (via SPIFFE Workload API): | |
| | - X.509-SVID (cert + private key) | |
| | - SVID bundle (trust roots) | |
| | | |
| | Performs: | |
| | - mTLS connection establishment | |
| | - EKM derivation | |
| | - Session-Binding Proof construction | |
| +--------------------------------------------+ |
| | mTLS (SVID-authenticated) |
+--------------------+--------------------------+---+
|
v
Remote Resource Server
The SPIFFE Workload API (typically a Unix domain socket) delivers the
current SVID and its private key to the workload process. The
workload uses this SVID as its mTLS client certificate and constructs
the Session-Binding Proof by signing with the corresponding private
key.
*SVID rotation*: When the SPIRE agent rotates the SVID (before
expiry), the workload receives the new SVID via the Workload API.
The workload SHOULD establish new mTLS connections using the new
SVID. Each new connection produces a new EKM; fresh proofs must be
constructed for any tokens presented on those connections.
Previously cached proofs, which reference the old connection's EKM,
remain valid only on connections established with the old SVID.
*Security note*: In this variant, the SVID private key resides in the
same process as the agent's application logic and LLM runtime. The
key-isolation benefit described in Appendix B does not apply. A
workload compromise may expose both the access token and the private
key. This variant is appropriate for environments where process-
Krishnan, et al. Expires 24 November 2026 [Page 33]
Internet-Draft TLS-Session-Bound-Tokens May 2026
level compromise is not the primary threat model, or where hardware-
backed key storage (e.g., PKCS#11, TPM) is enforced at the platform
level independently of the application.
C.2.2. With Sidecar (Recommended for Agentic AI)
In this variant, the SPIRE agent delivers the SVID exclusively to the
sidecar. The workload process never receives the SVID private key.
The sidecar holds the key, establishes all outbound mTLS connections,
and constructs Session-Binding Proofs transparently. This is the
pattern described in Appendix B, extended to SPIFFE/SPIRE
environments.
SPIRE Server
|
| (node + workload attestation)
|
SPIRE Agent (node-level)
|
| Workload API (UDS) — SVID delivered to sidecar only
|
+----------------------------------------------------+
| Pod / VM |
| |
| +------------------+ +-------------------+ |
| | AI Agent | | Security | |
| | (app logic, | | Sidecar | |
| | LLM runtime) | | (Envoy + SDS) | |
| | | | | |
| | Does NOT hold: | | Holds (via SDS): | |
| | - SVID | | - X.509-SVID | |
| | - private key | | - SVID private key| |
| | | | - EKM cache | |
| | | | - proof cache | |
| +--------+---------+ +--------+----------+ |
| | plaintext HTTP (loopback/UDS) |
| +------------>-----------+ |
| | mTLS (SVID) |
+------------------------------------+----------------+
|
v
Remote Resource Server
Envoy-based sidecars typically receive SVID updates from the SPIRE
agent via Secret Discovery Service (SDS). When the SVID rotates, the
sidecar receives the new certificate and key via SDS and uses it for
all subsequent outbound connections. The sidecar's connection
lifecycle management described in Appendix B applies directly: on
Krishnan, et al. Expires 24 November 2026 [Page 34]
Internet-Draft TLS-Session-Bound-Tokens May 2026
SVID rotation, a new mTLS connection to the resource server produces
a new EKM, and the sidecar constructs fresh proofs for any tokens
presented on that connection. Resource servers caching
(connection_id, ath) bindings are unaffected, as each cache entry is
keyed on the connection rather than the certificate.
C.3. Relationship to WIMSE WIT/WPT
SPIFFE X.509-SVIDs and WIMSE Workload Identity Tokens (WITs) serve
the same purpose — carrying attested workload identity — through
different credential formats. This specification is compatible with
both.
When WIMSE WIT/WPT is used alongside this specification in a
workload-to-workload flow, the two mechanisms address complementary
questions at different layers:
* *WIMSE WIT/WPT*: "Is this workload who it claims to be, and is
this the workload that holds the corresponding private key?"
(application-layer identity assertion and per-request proof of
possession)
* *This specification*: "Is this OAuth access token being presented
on the TLS connection that this workload authenticated?" (TLS-
channel-level token binding)
Together they close the complete chain: workload identity is asserted
and proven at the application layer (WIT/WPT), and the OAuth
authorization token is cryptographically bound to the specific
connection that workload established (this specification). Neither
mechanism alone provides both properties.
Authors' Addresses
Ram Krishnan
JPMorgan Chase & Co
Email: ramkri123@gmail.com
A Prasad
Oracle
Email: a.prasad@oracle.com
Diego R. Lopez
Telefonica
Email: diego.r.lopez@telefonica.com
Krishnan, et al. Expires 24 November 2026 [Page 35]
Internet-Draft TLS-Session-Bound-Tokens May 2026
Srinivasa Addepalli
Aryaka
Email: srinivasa.addepalli@aryaka.com
Krishnan, et al. Expires 24 November 2026 [Page 36]