]> TU Dresden, Germany

muhammad_usama.sardar@tu-dresden.de

CoreWeave

slava@dubeyko.com

Shanghai Guan An Information Technology Co., Ltd., China

bluedognull@gmail.com

Independent

aurestarnull@gmail.com

Shanghai Guan An Information Technology Co., Ltd., China

havan12050544@gmail.com

Switch

kaya.ercihan@switch.ch

Independent, Netherlands

evac.m.willems@proton.me

PHI-OMEGA

phiomega.runtime@gmail.com

HashCloak Inc and Stoffel Labs Inc, Canada

mikerah@hashcloak.com

EMILIA Protocol, Inc.

team@emiliaprotocol.ai

SEAT AI agents Intra-handshake attestation CVE-2026-33697 The draft aims to provide technical details of CVE-2026-33697, EUVD-2026-16488, and several GitHub Security Advisories (GHSAs) which provide substantial technical evidence of how intra-handshake (aka early) attestation fails in practice, even without physical access. Moreover, since continuous attestation is generally required, intra-handshake attestation adds unnecessary complexity. The results are backed by the research , and the artifacts in state-of-the-art formal analysis tool, ProVerif, under Apache-2.0 license for reproducibility and review, and have been acknowledged by the relevant stakeholders. Currently, there are two CVEs of CVSS 7.5, one GHSA of 9.0-10.0, two GHSAs of CVSS 9.1, one GHSA of CVSS 7.8, seven GHSAs of CVSS 7.4, and one GHSA of CVSS 6.3 published against the broader intra-handshake (aka early) attestation covering all layers of the ecosystem up to the application. The research papers on these are currently either under submission or being prepared for submission. The artifacts of these papers will be shared with the community under Apache-2.0 license for reproducibility and review. In our analysis, the remaining implementations of early attestation -- Edgeless Systems Contrast and Meta's AI -- remain vulnerable. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Source for this draft and an issue tracker can be found at .

Introduction presents a general approach to analyze the intra-handshake (aka early) attestation proposals, regardless of whether they are within the scope of SEAT charter or not. From a security perspective, one of the key decision factors is the candidate binding mechanism. Some binding mechanisms are within scope of SEAT charter and others are not. The artifacts are available in under Apache-2.0 license for reproducibility, extensibility and further research. A complementary paper presents the identity crisis in pre- and intra-handshake attestation. The formal analysis is available in under Apache-2.0 license for reproducibility and extensibility. Another complementary paper -- currently under submission -- performs a thorough formal analysis of the design options in intra-handshake attestation.

Overview presents the formal specification and analysis of the candidate binding mechanisms for binding in intra-handshake attestation for standardization for attested TLS protocols: Binding mechanisms, implementations and ProVerif artifacts

No.	Binding mechanism	Used in	Artifacts
1.	Client’s TLS nonce	-	binder1
2.	Client’s attestation nonce	-	binder2
3.	Early exporter	-	binder3
4.	Server’s public key	-	binder4
5.	Combination of #2 and #3	-	binder5
6.	Combination of #2 and #4	Edgeless Systems Contrast; Cocos AI v0.8.2; CCC Attestation SIG's adopted project intra-handshake attestation; Meta's AI updated spec	binder6
7.	Combination of #2, #3, and #4		binder7

We responsibly disclosed the vulnerability in intra-handshake attestation -- as noted in issued -- to the vendors, which resulted in of CVSS 7.5.

Modeling Other Binding Mechanisms The artifacts are quite flexible for modification and testing of different intra-handshake attestation binding mechanisms by simply changing single rdata parameter in the Client and Server processes. Folder aggregate contains all analyzed and proposed binding mechanisms in to select via comment and uncomment. Other folders contain one specific binding mechanism.

SEAT-Early-Attestation The draft is an extension of the provably vulnerable (and withdrawn) draft with the following two main changes from a formal perspective:

1. Binder has been updated
2. Post-handshake attestation part has been added for re-attestation

The current binder in does not prevent relay attacks as there is no shared secret in the binder. In addition to the formal analysis in , see for arguments why shared secret is necessary to prevent relay attacks. Post-handshake attestation part may prevent relay attacks, but then the additional complexity of intra-handshake attestation is unjustified.

Executive Summary of Current Status Severity is based on NIST metrics. Published CVEs/GHSAs for intra-handshake (aka early) attestation

CVSS	Severity	Number of Published CVEs/GHSAs
9.0-10.0	Critical	1
9.1	Critical	2
7.8	High	1
7.5	High	2
7.4	High	7
6.3	Medium	1

For TLS reference, Heartbleed was CVSS 7.5.

Published GHSAs/CVEs GHSAs/CVEs for intra-handshake (aka early) attestation and finders in (roughly) chronological order of publishing

GHSA/CVE	CVSS	Finders
	7.8	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.5	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.5	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar
	9.1	Muhammad Usama Sardar and Songbo Bu
	9.1	Muhammad Usama Sardar and Songbo Bu
	9.0-10.0	Markus Rudy; independently by Songbo Bu and Muhammad Usama Sardar
	7.4	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar, Viacheslav Dubeyko, and Jean-Marie Jacquet
	7.4	Muhammad Usama Sardar
	6.3	Muhammad Usama Sardar

Threat Model The threat model is explained in Sec. 6.1 of and Sec. 4 of . Beyond post-generation leakage of privEK considered in , the same adversary capability may arise from failures during key generation or entropy provisioning. Platform-attestation keys and workload-controlled TLS keys belong to distinct key-generation domains: for example, in AMD SEV-SNP the VCEK is derived by SNP firmware from chip-unique secrets and a TCB version, while several other platform secrets are specified as CSRNG-generated; by contrast, privEK and TLS (EC)DHE private values are typically generated by software executing inside the confidential VM using the guest OS or cryptographic-library random subsystem. Furthermore, SEV-SNP REPORT_DATA is supplied by the guest and incorporated into the signed attestation report without being interpreted by SNP firmware; consequently, valid Evidence can authenticate a binding value without attesting the entropy provenance, generation procedure, or exclusive possession of the corresponding private key. The LEK(privEK) capability should therefore also encompass predictable or repeated key generation caused by deficient entropy, cloned or rolled-back DRBG state, defective software or firmware, or malicious provisioning. CVE-2025-62626 provides a concrete manufacturer-layer fault model: affected AMD Zen 5 processors could return insufficiently random values from certain RDSEED forms while incorrectly signaling success. This does not establish compromise of the AMD-SP-internal CSRNG or of a specific attested-TLS implementation, but demonstrates that ideal-randomness assumptions can fail below the protocol layer; software dependencies such as OpenSSL's --with-rand-seed=rdcpu (OpenSSL 3.5.0 INSTALL.md), which can use RDSEED or RDRAND as CSPRNG seed input, illustrate a possible propagation path from hardware entropy interfaces to workload TLS key generation.

Low-Level Mapping of the System Model Figure 2 of provides a TEE-agnostic protocol-level abstraction. For a low-level view, the following table maps the abstract components to representative Intel TDX and AMD SEV-SNP implementations. Mapping of the abstract system model to representative CC implementations

Fig. 2 element	Intel TDX	AMD SEV-SNP
Physical Machine	TDX-capable Intel platform	SEV-SNP-capable AMD platform
CC Platform	CPU HW + TDX Module + attestation infrastructure	CPU HW + AMD-SP/SNP (system) firmware + RMP/SEV machinery
Quoting Agent	TD QE	AMD-SP / SNP attestation (VM) firmware
Confidential VM	Trust Domain (TD)	Part of SNP confidential VM
Network stack	Part of guest OS + TLS library inside TD	Part of guest OS + TLS library inside SNP guest
HSM/TPM	Secure element	Secure element
privAK	Attestation key of TD Quoting Enclave	VCEK/VLEK signing key
privEK	Workload/TLS-side ephemeral key	Workload/TLS-side ephemeral key
privLTK	Long-term key in secure element	Long-term key in secure element

The key material shown in the abstract model belongs to different implementation and trust domains. The following table provides a corresponding low-level view. Low-level implementation and key-generation domains

Component/key	Runs/lives where?	Type	Randomness/key source
TLS ECDHE	Inside network stack	Network stack	OS/library CSPRNG
privEK	Inside confidential VM	Guest software	OS/library CSPRNG
AK	Quoting Agent	Firmware/enclave/platform key hierarchy	Platform-specific
Memory-encryption key	CC Platform	Hardware/firmware managed	Platform RNG/KDF
REPORT_DATA	Created by Guest Software	Data binding	No independent entropy requirement

Per-VM memory-encryption key is used to encrypt confidential VM's RAM.

Detailed Vulnerability Disclosure Timeline and Public Acknowledgements by Affected Vendors Detailed vulnerability disclosure timeline and acknowledgements

Event	Date
Our initial responsible disclosure to vendor	07 Oct, 2025
Acknowledgement by vendor	14 Dec, 2025
Information to the IETF	11 Jan, 2026
Public announcement by vendor	27 Feb, 2026
Cocos AI published [Severity = HIGH (CVSS 7.8)]	23 March, 2026
CVE published [Severity = HIGH (CVSS 7.5)]	26 March, 2026
ENISA published EUVD [Severity = HIGH (CVSS 7.5)]	26 March, 2026
Acknowledgment by Privasys for rustls [Severity = HIGH (CVSS 7.5)]	9 July, 2026
Acknowledgment by Privasys for go [Severity = HIGH (CVSS 7.5)]	10 July, 2026
CCC implementation declared vulnerable to relay attacks	17 July, 2026
Vulnerable CCC implementation repo archived	22 July, 2026
Vulnerable draft withdrawn by authors	23 July, 2026
Edgeless Systems published [Severity = HIGH (CVSS 7.4)]	29 July, 2026
Cocos AI published [Severity = CRITICAL (CVSS 9.1)]	16 August, 2026
Cocos AI published [Severity = CRITICAL (CVSS 9.1)]	16 August, 2026
Edgeless Systems published [Severity = CRITICAL (CVSS 9.0-10.0)]	24 August, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	3 September, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	3 September, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	3 September, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	3 September, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	3 September, 2026
Privasys archived early attestation in rustls and moved to post-handshake attestation	4 September, 2026
Privasys archived early attestation in go and moved to post-handshake attestation	4 September, 2026
Privasys published [Severity = HIGH (CVSS 7.4)]	6 September, 2026
Privasys published [Severity = MEDIUM (CVSS 6.3)]	6 September, 2026

Neither the GHSAs nor the CVE has any dependency whatsoever on the considered threat model with WeakHash, WeakDH, or BadElement. They hold independent of those, i.e., with StrongHash and StrongDH and all good elements within a group.

EU ENISA European Union's ENISA has independently published with CVSS 7.5 to acknowledge this vulnerability.

Comparison with Other Vulnerabilities in Confidential Computing Literature Severity is based on NIST metrics. Comparison with other vulnerabilities in confidential computing literature

Vulnerability	CVE	CVSS	Severity
wiretap.fail	No CVE (Intel and AMD announcements)	-	None
TEE.fail	No CVE	-	None
TDXdown	Intel	2.5	Low
Staleus	CVE-2025-54509	4.0	Medium
BreakFAST	CVE-2025-61972	4.2	Medium
BadRAM	AMD	5.3	Medium
BreakFAST	CVE-2025-61971	5.9	Medium
Fabricked	CVE-2025-54510	5.9	Medium
Intra-handshake.fail		7.5	High

The comparison of the above with CVSS 7.5 for indicates that attested TLS is not mature yet compared to the rest of the confidential computing stack, and is currently one of the weakest links in the ecosystem.

More CVEs Further formal analysis has led to the following potential CVEs for intra-handshake (aka early) attestation (currently under review and disclosure): Expected CVEs for intra-handshake (aka early) attestation under review and disclosure

CVSS	Severity	Number of CVEs
9.0-10.0	Critical	1 (confirmed by developers)
9.8	Critical	1
9.1	Critical	2 (confirmed by developers)
8.7	High	1
7.5	High	5
7.4	High	9 (5 confirmed by developers)
6.3	Medium	7

These are preliminary estimates of scores, not final assigned score. They are still under review.

Vulnerable Implementations As demonstrated in and , at least the following intra-handshake implementations are vulnerable:

• Meta's AI: and [Severity = HIGH (CVSS 7.5)]
• Edgeless Systems Contrast: [Severity = HIGH (CVSS 7.4)]

If you are aware of any other intra-handshake attestation implementation, please let us know so that we can check and responsibly disclose the vulnerabilities to them.

Archived/Mitigated Implementations The following intra-handshake implementations were vulnerable and have been archived or moved to post-handshake attestation:

• CCC Attestation SIG's adopted project intra-handshake attestation: declared vulnerable to relay attacks and archived
• Cocos AI <= v0.8.2: [Severity = HIGH (CVSS 7.8)], and [Severity = HIGH (CVSS 7.5)]; migrated to post-handshake attestation since v0.9.0
• Privasys rustls <= privasys-v0.2.0: [Severity = HIGH (CVSS 7.4)], archived and Privasys migrated to post-handshake attestation
• Pirvasys go <= privasys-v0.3.0-go1.26.5: [Severity = HIGH (CVSS 7.4)], archived and Privasys migrated to post-handshake attestation

Vulnerable Protocol Specifications At least the following protocol specifications with intra-handshake attestation path are vulnerable to and :

• : symbolic proof of insecurity; withdrawn after the CVE
• : symbolic and (paper-and-pen-based) computational proof of insecurity (originally done for -04 and applies also to -06)
  • As a SEAT WG participant pointed out, please note that both and contain a link to that contains a link to that contains the G3 property (cf. ) that this draft does not satisfy.
  • Some WG participants successfully reproduced the vulnerability by substituting the right value of rdata in the shared formal model that led to the CVE.
  • An informal reasoning is that binder is not directly derived from any shared secret in this draft.
  • Unnecessary complexity is itself a security concern
• : symbolic proof of insecurity
  • violates G3 property in our analysis
  • unnecessary complexity is itself a security concern

Binding Levels

1. DH shared secret (gxy) used as shared secret between client and server
2. Handshake traffic key (htsc) used for encryption of handshake messages
3. Application traffic key (atsc) used for encryption of application data

Please see Sec. 6.2 of for details.

Security Properties (Correlation Goals) We consider TLS Server as RATS Attester, which is typical in confidential computing.

1. Correlation of Evidence to a DH Shared Secret (G1)
2. Correlation of Evidence to Client’s Handshake Traffic Key (G2)
3. Correlation of Evidence to Client’s Application Traffic Key (G3)

Please see Sec. 6.3 of for details.

Main Results

• All analyzed binding mechanisms and the corresponding implementations of intra-handshake attestation are vulnerable to relay attacks.
• Early exporter helps achieve level 1 binding.
• Our proposed mechanism helps achieve level 2 binding.
• It may not be possible to achieve level 3 in intra-handshake attestation alone without additional assumptions.

Main results

Property	Mechanism #1,2,4,6	Mechanism #3,5,7	Proposed mechanism
G1 : Correlation of Evidence to gxy	❌	✅	✅
G2 : Correlation of Evidence to kch	❌	❌	✅
G3 : Correlation of Evidence to kc	❌	❌	❌

Please see Sec. 7.1 and Figure 5 of for details of attacks.

Expected Results Expected results

No.	Binding mechanism	Artifacts	Expected results
1.	Client’s TLS nonce	binder1	binder1
2.	Client’s attestation nonce	binder2	binder2
3.	Early exporter	binder3	binder3
4.	Server’s public key	binder4	binder4
5.	Combination of #2 and #3	binder5	binder5
6.	Combination of #2 and #4	binder6	binder6
7.	Combination of #2, #3, and #4	binder7	binder7
8.	Proposed	proposal	proposal

Implications of Findings

Implications of Findings for IETF SEAT WG

• We believe post-handshake attestation alone, such as draft-fossati-seat-expat, can achieve level 3 binding.
• The research suggests that recent hybrid proposals (combination of intra-handshake attestation and post-handshake attestation) and may add unnecessary complexity of intra-handshake attestation without adding any security benefit compared to post-handshake attestation alone, such as draft-fossati-seat-expat. We are not aware of any security property that hybrid proposals can achieve that post-handshake attestation alone cannot achieve.
• As demonstrated by our symbolic analysis using ProVerif, the protocol specifications and remain vulnerable to CVE-2026-33697. We have also proved that and violate the security theorems in the computational model.

Implications of Findings for IETF LAKE WG

• Similar problems occur for protocol specification lake-ra.

Implications of Findings for IETF TLS WG

• is vulnerable to . Thankfully, the authors have withdrawn .
• Remote attestation within the handshake is very dangerous, since to our knowledge, it is one of the highest scored published vulnerabilities in confidential computing literature (see ). For reference, Heartbleed was 7.5 CVSS.

Implications of Findings for Agent2Agent The findings of published CVEs/GHSAs up to 9.1 (presented in ) show that intra-handshake attestation can introduce significant security risks for AI agents when relied upon as a security mechanism. Attestation can provide evidence about an agent’s technical state, but such evidence should not be equated with governability. For a relying party, governability also depends on whether the agent’s identity, authority and permissions remain aligned with the intended interaction, whether responsibility for its actions can be attributed, and whether meaningful intervention remains possible. The findings in this draft reinforce that distinction by showing that even the binding between attestation evidence and the intended session can fail. Successful attestation should therefore be treated as one input into governance, rather than as sufficient evidence that an AI agent remains under effective control.

Technical Details

Tool We use state-of-the-art symbolic security analysis tool ProVerif for the specification of the protocols.

Modeling The formal model uses the fixed version of diversion attacks in intra-handshake attestation from our previous work as the starting point to focus on relay attacks in intra-handshake attestation in this work. The rationale is that we consider it more useful to show the added value of this contribution to the community by using the fixed version of diversion attacks in intra-handshake attestation as the baseline, rather than showing the same diversion attacks from , and the discovered CVE () -- which the previous analysis could not find -- practically demonstrates the added value. This modeling choice makes it clear that even with the diversion attacks fixed, high-severity relay attacks would still remain in intra-handshake attestation. Note: Similar to the fixed version of diversion attacks in intra-handshake attestation from our previous work, we model non-PSK-based handshake. From :

• For modeling TLS 1.3, we consider handshakes based on Diffie-Hellman over either finite fields or elliptic curves, represented as (EC)DHE. This is because we are unaware of any publicly available specification or implementation of attested TLS with PSK-based handshakes.

While it would be nice to model PSK-based handshake, the rationale is that the correlation properties studied in this work do not necessarily require it. Note: The artifacts consider the case of server authentication only, as client authentication is optional in TLS 1.3. No claims are made about other configurations.

Properties Properties in are complemetary to properties in . Sec. 8 of mentions:

• We emphasize that both diversion and relay attacks are orthogonal and thus the two works are complementary.

Technical Vulnerability Report Technical vulnerability report is available at . It is accepted for publication at ESORICS 2026.

Vulnerabilities Sec. 7.1 of presents the technical details with abstract attack traces of the vulnerabilities.

Mitigation Sec. 7.2 of presents the technical details of the proposed mitigation.

Artifacts Artifacts are available at under Apache-2.0 License.

Media Coverage Several media professionals and bloggers have covered the vulnerabilities to protect the community from the harm of intra-handshake attestation.

• The Register
• (Japanese) BlackHatNewsTokyo
• (Several languages) Hackernoon
• Apple podcast
• Information Security News
• TheNextGenTechInsider
• DailySecurityReview
• SC World
• 01 Quantum
• (Russian) Security Lab
• (German) blogspan
• (Chinese) Sina
• data4biz
• (Russian) ITSec
• (Chinese) smzdm
• (Chinese) donews
• (Chinese) ifeng
• dugganusa
• dugganusa repo
• spoitus
• lavx news
• sohu
• (Persian) news.ditty
• (Russian) LiMP VPN
• daily.dev
• warden
• GCVE.eu
• vuln.lu
• coderlegion
• Anjuna Security
• Privasys
• Caution
• freenode
• (Chinese) csdn
• osintsights
• (Turkish) hardwaremania
• akber
• ad-hoc news
• AIMultiple

Security Researchers Several credible security researchers, such as the following, have publicly attested to it.

• Michael Pak
• Rodrigo Branco
• Bart Preneel
• Thorsten Strufe

Germany's BSI Germany's Federal Office for Information Security (Bundesamt für Sicherheit in der Informationstechnik) has attested to it. Carina Hilt, deputy press spokesperson at BSI, told The Register: CC refers to Confidential Computing, and attested TLS is the core trust mechanism of CC.

Reviews

Conference Reviews has been peer-reviewed and accepted for publication at ESORICS 2026.

IETF/IRTF Several participants of the IETF/IRTF have attested to the results by independently reproducing the results and reviewing the code. Some of the participants have independently reproduced the results by developing their own formal models and a proof-of-concept implementation of the vulnerabilities. Some of the messages are mentioned below (excluding the messages of authors of ):

• https://mailarchive.ietf.org/arch/msg/seat/B7F1Dj_rjs8I0Kg3yCp3Rap0XeE/
• https://mailarchive.ietf.org/arch/msg/seat/aEV9dUFotAQzHndk23qBcwBT3as/
• https://mailarchive.ietf.org/arch/msg/seat/3Hv0E1sfXsvyBtl6AgY8j-SHHiw/
• https://mailarchive.ietf.org/arch/msg/seat/5LJ6i9svomnhpyPHWPejM7fMmXQ/
• https://mailarchive.ietf.org/arch/msg/seat/V_YqGUY3fEwaFwwyfA9DshpHet0/
• https://mailarchive.ietf.org/arch/msg/seat/JF_cwmHHEbrJ_W5V6yEetWWnii4/
• https://mailarchive.ietf.org/arch/msg/seat/P_CYTycg0KG7cbKauFA-kVgX2jo/
• https://mailarchive.ietf.org/arch/msg/seat/ZJjJXpYwZ5nCVmz_W4FK6XiFEY4/
• https://mailarchive.ietf.org/arch/msg/seat/4so3LxHOOXHS1wnvhuoeWWgeCHk/
• https://mailarchive.ietf.org/arch/msg/seat/Q6Jmc58v0c1lDV3ujIY0AX_ofGA/
• https://mailarchive.ietf.org/arch/msg/seat/n4Me5QPCvwhxcEJndWePyishcoo/
• https://mailarchive.ietf.org/arch/msg/seat/beRzNNvwMifkRfJPfxecGoHpTDs/
• https://mailarchive.ietf.org/arch/msg/cfrg/U5YHd91lYjiqCTt9BZyVDNFeUpM/
• https://mailarchive.ietf.org/arch/msg/seat/aFCo4BMRDSUynvN9AQJatPjnXag/
• https://mailarchive.ietf.org/arch/msg/seat/wb_Ys9MZd9u9oM2Bk-8tv7fvXGg/
• https://mailarchive.ietf.org/arch/msg/seat/ov8f-7cZKK5RZ-Mmjc6IhVSB-Fk/
• https://mailarchive.ietf.org/arch/msg/seat/2uUuaD1DygjNDM4GT_-rYbwTiJk/
• https://mailarchive.ietf.org/arch/msg/seat/pB39abN1QrH4_ATM_E78vxPTuxk/
• https://mailarchive.ietf.org/arch/msg/seat/PxKCxMHe-SAiR9uhOOllrK4mUA4/
• https://mailarchive.ietf.org/arch/msg/seat/T1xupUBwqYEBSHCTXgSHXZtdqz8/
• https://mailarchive.ietf.org/arch/msg/seat/hRw46FwgmVdi9fqZm2fjKbln_IA/
• https://mailarchive.ietf.org/arch/msg/seat/UG7yE_klmRSxNy2HX6fzuFonDjM/
• https://mailarchive.ietf.org/arch/msg/seat/2hpeIldeFfE6o9q6L9Vkt00ACKA/
• https://mailarchive.ietf.org/arch/msg/seat/gc2ij0vboehS_-v10-SNslxaZC0/
• https://mailarchive.ietf.org/arch/msg/seat/oO4mAfq5HJZptDDNrSnd7zDdX18/
• https://mailarchive.ietf.org/arch/msg/seat/XuJc_yEJPCMIuYcv2OM7XDogRCU/
• https://mailarchive.ietf.org/arch/msg/seat/nVHlnbFIEh-cPQeMuDVOqx5YvWQ/
• https://mailarchive.ietf.org/arch/msg/seat/xVU3C7qUOngcip7B4ZO5MJUT9Xg/
• https://mailarchive.ietf.org/arch/msg/seat/1gCcPw-7NopDRzzBzA3dFIgo3Rs/
• https://mailarchive.ietf.org/arch/msg/seat/t8aobzB374lWiLzrVrORY7kGYyQ/
• https://mailarchive.ietf.org/arch/msg/seat/m3UyB6XLQzxaucejE_o8Pn41uSI/
• https://mailarchive.ietf.org/arch/msg/seat/gqHqcbbKva_oGE-jEDZu243gf-4/
• https://mailarchive.ietf.org/arch/msg/seat/QD8QB1WVL-toNovGQ2Tk6DmmeEM/
• https://mailarchive.ietf.org/arch/msg/seat/vXN2pifZ5GXcC1xLwSfLCnUcFUE/
• https://mailarchive.ietf.org/arch/msg/seat/MGFXinb85XSaLkqjBEhmBZC7PcI/
• https://mailarchive.ietf.org/arch/msg/seat/js9VI4PB8yYmhg2ObaZB1a22fL4/
• https://mailarchive.ietf.org/arch/msg/seat/0RzORzX_VdlY5UQ_MWMmxZlnrjs/
• https://mailarchive.ietf.org/arch/msg/seat/W3MH1BDSUbm1WUPxGQihaIc1zTk/
• https://mailarchive.ietf.org/arch/msg/seat/2_aGylmFHoLmqN7BBcYVoH-rNJk/
• https://mailarchive.ietf.org/arch/msg/seat/7SYSuB83Kmr9qCb1V1F94n9W33U/
• https://mailarchive.ietf.org/arch/msg/seat/0SWfg2YNEAOtJQ7Zsf1xl4O-AOo/
• https://mailarchive.ietf.org/arch/msg/seat/1mfNw-bw8KsdJbl4saL99Fz4iec/
• https://mailarchive.ietf.org/arch/msg/seat/VyifG8zP5aworb_S1NR9FEUEXW0/
• https://mailarchive.ietf.org/arch/msg/seat/CYwvM75z6rTId2A3ZZvZJmHxOig/
• https://mailarchive.ietf.org/arch/msg/ufmrg/29xFZX5C4oSGkpZAvXT_7YLW2Vc/
• https://mailarchive.ietf.org/arch/msg/seat/u1HxYW9cJfVpi3Cf9q06ehwpYGE/
• https://mailarchive.ietf.org/arch/msg/seat/SG_A0016a-KMnXAkGtMUxokZmjc/
• https://mailarchive.ietf.org/arch/msg/seat/3w7-OW2CAVr0-QBz97eAMxB_nMI/
• https://mailarchive.ietf.org/arch/msg/seat/UnybcafvQ2D-IhUfTV228WQFNhA/
• https://mailarchive.ietf.org/arch/msg/seat/rmVNeFbjax26l31n5pitHIxOQkk/
• https://mailarchive.ietf.org/arch/msg/seat/DghJdG3ysbPFKMQe8czz-tcIMq0/
• https://mailarchive.ietf.org/arch/msg/seat/rZLacid2wnEtaJwSbiIIft3T0FI/
• https://mailarchive.ietf.org/arch/msg/seat/_kEBODNsTWjgadb5xnlj86dvhcs/
• https://mailarchive.ietf.org/arch/msg/seat/qP3XC0MarFFA3SMbBpWWJtxACNA/
• https://mailarchive.ietf.org/arch/msg/seat/kkjQhi4yvJ_iAwYrPw1crFh-m-0/
• https://mailarchive.ietf.org/arch/msg/seat/vBkdKtKzTt4F91VprfKIndgmT2o/
• https://mailarchive.ietf.org/arch/msg/seat/Huu_AFu11BTrdxK3I8hmw2jjp8Q/
• https://mailarchive.ietf.org/arch/msg/seat/oOnioxkB__QZIvhFn5naW6jIXzg/
• https://mailarchive.ietf.org/arch/msg/seat/iWsCCAl8YZ-pOTA7siNUGsfliHQ/
• https://mailarchive.ietf.org/arch/msg/seat/-HGPUR5CvuVWcOAg37cSxwoATm0/
• https://mailarchive.ietf.org/arch/msg/seat/LnLYE7bGQOmCxVXq6stiOtKwc1s/
• https://mailarchive.ietf.org/arch/msg/seat/o_bIJhOdB4j1g0nczxPZwFXtCo8/
• https://mailarchive.ietf.org/arch/msg/seat/hy4qVQJQGR82-bskQ_UGI6iel1Y/
• https://mailarchive.ietf.org/arch/msg/seat/3J3s_YFnf9IQ87Tv2c1q4f36xKQ/
• https://mailarchive.ietf.org/arch/msg/seat/JWKMYY1YG1E2iS_HyQ4rDmOsDGw/
• https://mailarchive.ietf.org/arch/msg/seat/rK1nDSewAbVL_weOp98knYZcg6s/
• https://mailarchive.ietf.org/arch/msg/seat/Wjuz0fIj8tjYocUmiZZXcSwwFHw/
• https://mailarchive.ietf.org/arch/msg/seat/SYiV4KZNr20re6QkGmyWS3pPteA/
• https://mailarchive.ietf.org/arch/msg/seat/ZYgxm1ibt6p4dL7xF1YNdl0XSpc/
• https://mailarchive.ietf.org/arch/msg/seat/6LKgOp22YRxGTYb-i-BxiMGzMW4/
• https://mailarchive.ietf.org/arch/msg/seat/3oHcKPtXLfBUYCZoN1nnO6e1F-s/
• https://mailarchive.ietf.org/arch/msg/seat/aSybH9ihnNjN7SjdhhY_XtxhMtc/
• https://mailarchive.ietf.org/arch/msg/seat/d_G8r7LMGjA45BPexZwsfmmAtJY/
• https://mailarchive.ietf.org/arch/msg/seat/u-za86_YJ0spwBXBwTVQzwiOCrU/
• https://mailarchive.ietf.org/arch/msg/seat/P4IMdvCx8lSEKrCiJIjbpBCRr4g/
• https://mailarchive.ietf.org/arch/msg/seat/-AO9yFJoflV1wDwwch45dmqkmyo/
• https://mailarchive.ietf.org/arch/msg/seat/H0LqHfBasQml69Y-9Zl_njfsE8s/
• https://mailarchive.ietf.org/arch/msg/seat/3hOGlYyc_yntmvT0tjYxldlwaxM/
• https://mailarchive.ietf.org/arch/msg/seat/JbwL9cdl6fiP0vBGgASUUDmaPy8/
• https://mailarchive.ietf.org/arch/msg/seat/TtewHbMGKBQOKD2sqrgTrnpCOkI/
• https://mailarchive.ietf.org/arch/msg/seat/UsIj6o7wf4hX_uCL51TCTv-wFxU/
• https://mailarchive.ietf.org/arch/msg/seat/a6c8D6PBDRe0x4DAqXM3t4EPc4c/
• https://mailarchive.ietf.org/arch/msg/seat/prB537Jht2kELVCSrTRktjbp7Y4/
• https://mailarchive.ietf.org/arch/msg/seat/RPnKYfUCD_MRw3hnA00zg684yGM/
• https://mailarchive.ietf.org/arch/msg/seat/nMH_0sLLU5MekoEnzWKJnIJmaSk/
• https://mailarchive.ietf.org/arch/msg/seat/GJCA31mgAehlgFPRu_yHA10lKPI/
• https://mailarchive.ietf.org/arch/msg/seat/9f-21JMK6s1Pdcob6mPo06rNwmQ/
• https://mailarchive.ietf.org/arch/msg/seat/8qq_GFT391IEbGZZQUtYMONX9U0/

Main Questions In short, five main questions have been raised by WG participants in support of our work:

• What security property hybrid (intra- + post-handshake attestation) provides that post-handshake attestation alone cannot provide?
• Since continuous attestation is required in most use cases, how is additional complexity of intra-handshake attestation justified? Use cases with one-time attestation can be covered by doing attestation round immediately after Connection Establishment Time: see reference.
• What is the benefit of doing signatures of remote attestation within the handshake (as this latency can be exploited)? We add that verification of signatures is also time consuming, which can be exploited too. See reference.
• How evidence is bound to the secure channel without involving any shared secret? See .
• How does a verifying relying party get the legitimate PIIDs and CHIP_IDs?

Guidance Text

• Evidence MUST be bound to the secure channel. Failure to do so results in relay attacks , , .
• Verifier MUST have access to legitimate hardware identifiers of the Attester. Failure to do so results in relay attacks .
• Verifier MUST carefully check the binding. Failure to do so results in relay attacks , .
• Binder MUST contain shared secrets. Failure to do so results in relay attacks , , , , , , .

Researchers outside of IETF/IRTF Some researchers have approached us confirming the proof-of-concept of the vulnerabilities in intra-handshake attestation. More information will be added once their pre-prints/papers are public.

Security Considerations All of this document is about the insecurity of intra-handshake (aka early) attestation. By no means should the vendors mentioned in this draft be considered less secure than any other vendors implementing intra-handshake attestation solutions. In particular, those who have closed-source implementations are most likely more vulnerable than the open-source ones, since the former cannot easily be reviewed by the security community. Even extensive security reviews -- of closed-source implementations -- by cybersecurity firms often do not perform formal analysis, and thus such reviews may miss corner cases and subtle vulnerabilities.

Ethical Considerations We (i.e., the super set of all authors involved in this research, including but not limited to Muhammad Usama Sardar, Mariam Moustafa, Tuomas Aura, Viacheslav Dubeyko, Jean-Marie Jacquet, Songbo Bu, Chengxin Huang, Haowen Song, Kaya Ercihan, Massimiliano Brighindi, and Iman Schrock) are ethical researchers aiming to protect the community from the potential harm caused by the exploitability of the vulnerabilities in intra-handshake attestation. We have responsibly disclosed the vulnerabilities to the respective developers and maintainers following their respective disclosure processes and provided them our proposed mitigations and requested them to take rapid action. We have released only the formal analysis for published CVE. To minimize exploit in the wild, we have not publicly released the proof-of-concept exploit code. We have not retrieved any real data from any real system. We have not released any key to any public forum or to any person.

Evidence of Explanation of Vulnerabilities to the Authors of Vulnerable Drafts To the best of our abilities, knowledge, and understanding, we have tried to explain the vulnerabilities to the authors of vulnerable drafts , , and first privately in several meetings and then later on publicly for at least half a year at several forums, including but not limited to CCC Attestation SIG and IETF/IRTF. Please see the (non-exhaustive list of) recordings and the archives below. We sincerely thank the authors of for withdrawing their draft to protect further exploits mentioned in .

Recordings Evidence of several explanations of vulnerabilities to the authors of vulnerable drafts

Event/Host	Venue	Date(s)	Evidence
Linux Plumbers Conference 2026	Prague, Czechia	5-7 Oct, 2026	slides, video
GA4GH 14th Plenary Meeting	Singapore	28 Sept-2 Oct, 2026	slides, video
ESORICS 2026	Rome, Italy	14-18 Sept, 2026	slides
IETF RATS Interim meeting	Virtual	14 Sept, 2026	slides, video
Hackathon @ RIOT Summit 2026	Grenoble, France	4 September, 2026	topic synopsis
RIOT Summit 2026	Grenoble, France	2-4 September, 2026	abstract, slides, video
Data Security Work Stream (DSWS) at the Global Alliance for Genomics and Health (GA4GH)	Virtual	24 Aug, 2026	slides, video
Confidential AI Public Side Meeting @ IETF 126	Vienna, Austria	21 July, 2026	plan, slides
SEAT @ IETF 126	Vienna, Austria	21 July, 2026	slides, video
IETF 126 Hackdemo Happy Hour	Vienna, Austria	20 July, 2026	Hackathon project, demo
Confidential Computing Public Side Meeting @ IETF 126	Vienna, Austria	20 July, 2026	plan, slides
HotRFC @ IETF 126	Vienna, Austria	19 July, 2026	slides, video
IETF 126 Hackathon	Vienna, Austria	19 July, 2026	Hackathon project, slides, video
IEPG @ IETF 126	Vienna, Austria	19 July, 2026	slides, video
Workshop @ Dresden Science Night 2026	Dresden	26 June, 2026	demo
Output 2026	Dresden	25 June, 2026	demo
Confidential Computing Summit 2026 (presented by Jens Albers)	San Francisco, USA	23-24 June, 2026	poster
Confidential Containers Community Meeting @ Cloud Native Computing Foundation	Virtual	30 April, 2026	slides, video around timestamp 00:27:00
GIF Project showcase @ GA4GH April Connect 2026	Montreal, Canada (virtual)	17 April, 2026	slides, video, report
NSA Symposium on Hot Topics in the Science of Security (HotSoS) 2026	Virtual	16 April, 2026	abstract, slides, video
PET-CON 2026.1: 15th Privacy Enhancing Techniques Convention	Karlsruhe, Germany	16-17 April, 2026	slides, poster
GTMFS 2026: Annual Meeting of the WG "Formal Methods in Security"	Luz-Saint-Sauveur, France	24-26 Mar, 2026	slides
CFRG @ IETF 125	Shenzhen, China (virtual)	19 Mar, 2026	slides, video
SEAT @ IETF 125 (relay)	Shenzhen, China (virtual)	17 Mar, 2026	slides, video
Side meeting @ IETF 125	Shenzhen, China (virtual)	16 Mar, 2026	slides
LAKE @ IETF 125	Shenzhen, China (virtual)	16 Mar, 2026	slides, video
HotRFC @ IETF 125	Shenzhen, China (virtual)	15 Mar, 2026	slides, video
IETF 125 Hackathon	Shenzhen, China (virtual)	14-15 Mar, 2026	Hackathon project, slides, video
CCC Attestation SIG	Virtual	10 Feb, 2026	slides; video
IETF RATS Interim meeting	Virtual	9 Feb, 2026	slides, video
Confidential Computing devroom at FOSDEM 2026	Brussels, Belgium	31 Jan-1 Feb, 2026	abstract, slides, video
CCC Attestation SIG	Virtual	27 Jan, 2026	slides; video
CCC Attestation SIG	Virtual	13 Jan, 2026	slides; video
CCC Attestation SIG	Virtual	16 Dec, 2025	slides; video
CCC Attestation SIG	Virtual	2 Dec, 2025	slides; video

Archives Since January, we have publicly informed the authors of vulnerable drafts , , and and shared our results with the community for review and to raise awareness on high-severity vulnerabilities and apply appropriate mitigations for the safety of their users:

IETF

• SEAT WG
• RATS WG
• TLS WG
• LAKE WG
• SAAG
• Practical Cybersecurity list
• Agent2agent list thread1 and thread2
• DSMC list
• Hackathon
• 126attendees

IRTF

• UFMRG: thread1 and thread2
• CFRG thread1 and thread2
• DINRG

CCC

• Attestation SIG: thread1 and thread2
• TAC: thread1 and thread2

OCP

• OCP Security: message1, message2, message3 and message4

If you know any other relevant mailing list that we should inform for protection of users, please let us know.

Contributions Contributions to the draft are welcome at https://github.com/muhammad-usama-sardar/intra-handshake-fail.

IANA Considerations This document has no IANA actions.

References Normative References CVE ENISA TU Dresden, Dresden, Germany Aalto University, Espoo, Finland Aalto University, Espoo, Finland ACM IEEE Informative References Intuit Arm Limited Arm Limited Linaro Nokia The TLS handshake protocol allows authentication of one or both peers using static, long-term credentials. In some cases, it is also desirable to ensure that the peer runtime environment is in a secure state. Such an assurance can be achieved using remote attestation which is a process by which an entity produces Evidence about itself that another party can use to appraise whether that entity is found in a secure state. This document describes a TLS extension that enables the negotiation and binding of the TLS authentication key to a remote attestation session. This enables an entity capable of producing attestation Evidence, such as a confidential workload running in a Trusted Execution Environment (TEE), or an IoT device that is trying to authenticate itself to a network access point, to present a more comprehensive set of security metrics to its peer. This extension has been designed to allow the peers to use any attestation technology, in any remote attestation topology, and to use them mutually. Intuit Arm Limited Arm Limited Linaro Nokia The TLS handshake protocol allows authentication of one or both peers using static, long-term credentials. In some cases, it is also desirable to ensure that the peer runtime environment is in a secure state. Such an assurance can be achieved using remote attestation which is a process by which an entity produces Evidence about itself that another party can use to appraise whether that entity is found in a secure state. This document describes a series of TLS extensions that enable the binding of the TLS authentication key to a remote attestation session. This enables an entity capable of producing attestation Evidence, such as a confidential workload running in a Trusted Execution Environment (TEE), or an IoT device that is trying to authenticate itself to a network access point, to present a more comprehensive set of security metrics to its peer. These extensions have been designed to allow the peers to use any attestation technology, in any remote attestation topology, and to use them mutually. Intuit Arm Limited Arm Limited Arm Limited Huawei Linaro The TLS handshake protocol allows authentication of one or both peers using static, long-term credentials. In some cases, it is also desirable to ensure that the peer runtime environment is in a secure state. Such an assurance can be achieved using attestation which is a process by which an entity produces evidence about itself that another party can use to appraise whether that entity is found in a secure state. This document describes a series of protocol extensions to the TLS 1.3 handshake that enables the binding of the TLS authentication key to a remote attestation session. This enables an entity capable of producing attestation evidence, such as a confidential workload running in a Trusted Execution Environment (TEE), or an IoT device that is trying to authenticate itself to a network access point, to present a more comprehensive set of security metrics to its peer. These extensions have been designed to allow the peers to use any attestation technology, in any remote attestation topology, and mutually. Intuit Arm Limited Arm Limited Arm Limited Huawei Linaro The TLS handshake protocol allows authentication of one or both peers using static, long-term credentials. In some cases, it is also desirable to ensure that the peer runtime environment is in a secure state. Such an assurance can be achieved using attestation which is a process by which an entity produces evidence about itself that another party can use to appraise whether that entity is found in a secure state. This document describes a series of protocol extensions to the TLS 1.3 handshake that enables the binding of the TLS authentication key to a remote attestation session. This enables an entity capable of producing attestation evidence, such as a confidential workload running in a Trusted Execution Environment (TEE), or an IoT device that is trying to authenticate itself to a network access point, to present a more comprehensive set of security metrics to its peer. These extensions have been designed to allow the peers to use any attestation technology, in any remote attestation topology, and mutually. Intuit Arm Limited Arm Limited Arm Limited Huawei Linaro This draft has been withdrawn. About This Document This note is to be removed before publishing as an RFC. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-fossati-tls-attestation/. Source for this draft and an issue tracker can be found at https://github.com/yaronf/draft-tls-attestation. Independent This document describes FACTS (Factor-based Attestation and Credential Transport Scheme) over TLS 1.3. Conceptually acting as "multi-factor authentication" for machine identities, factor-based attestation derives session trust from multiple independent cryptographic inputs rather than a single point of failure. Specifically, it utilizes a dual-key scheme that binds identity to attestation evidence through the use of key encapsulation material keys (KEM) and traditional identity signing keys (IK), establishing per-session freshness.

Acknowledgments Acknowledgment does not necessarily imply attestation. It implies that the authors found the feedback and discussion useful in improving the formal analysis, the corresponding paper, or this draft. This draft benefits from several years of research on attested TLS, in particular some of the recent works mentioned below: We wish to express our sincere appreciation to the following for their review of our latest work:

• Bertrand Foing
• Sammy Kerata Oina
• Drasko Draskovic
• Markus Rudy
• Rebekah Overdorf
• Tobias Pulls

Intra-handshake.fail We would like to thank our co-author of paper for his valuable contributions:

• Jean-Marie Jacquet

We also gratefully acknowledge the following for insightful discussions and helpful reviews on :

• Eric Rescorla
• Juho Forsén
• Markus Rudy
• Mariam Moustafa
• Bruno Blanchet
• Steve Kremer
• Tjaden Hess
• Martin Thomson
• Yuning Jiang
• Pavel Nikonorov
• Casey Wilson
• Anonymous ESORICS 2026 reviewers
• Danko Miladinovic
• John Preuß Mattsson
• Britta Hale
• Werner Staub
• Songbo Bu
• Haowen Song
• Chengxin Huang
• Steve Luo
• Andrew Miller
• Kubilay Ahmet Küçük
• Iman Schrock
• Sophie Schmieg
• Davyd Okaianchenko
• Alistair Woodman
• Göran Selander
• Tom Sato
• Jakub Maria Plutowski
• Martin Friedrich
• Patrick Duggan
• Deb Cooley

Identity Crisis We would like to thank our co-authors of complementary paper for their valuable contributions:

• Mariam Moustafa
• Tuomas Aura

We also gratefully acknowledge the following for insightful discussions and helpful feedback:

• Ionut Mihalcea
• Jean-Marie Jacquet
• Thomas Fossati
• Eric Rescorla
• Hannes Tschofenig
• Yaron Sheffer
• Laurence Lundblade
• Giridhar Mandyam
• Christopher Patton
• Jonathan Hoyland
• Richard Barnes

refTLS We sincerely thank the following for the foundational formal model of draft 20 of TLS 1.3 in their work that we have used as the foundation of all of this work:

• Karthikeyan Bhargavan
• Bruno Blanchet
• Nadim Kobeissi

General Several others at the IETF, IRTF, CCC, and GA4GH have contributed by providing feedback over the years. A non-exhaustive list of contributors is here. Muhammad Usama Sardar is funded by German Research Foundation ("Deutsche Forschungsgemeinschaft.")