← All intelligence

web · November 6, 2026

The Trail of Bits Blog

blog.trailofbits.com · Read at source ↗

Since 2012, Trail of Bits has helped secure some of the world's most targeted organizations and products. We combine high-­end security research with a real­ world attacker mentality to reduce risk and fortify code.

Trail of Bits’ latest blog entries cover blockchain security, AI-assisted software repair, cryptography and testing tools, including DAML support for its mutation-testing engine and a joint open-source security initiative with OpenAI.

Wallet and blockchain security

Trail of Bits outlines four measures to strengthen wallet security after Binance: restrict signing access, control delegation, limit potential losses and detect changes in wallet authority.

The firm also reports finding and disclosing a vulnerability in Provenance Blockchain, a public proof-of-stake chain built on Cosmos SDK. The flaw allows any user to give themselves administrative control over marker accounts without owning any tokens.

Another entry describes seven recurring failure patterns in application and hook code. These include absent caller checks and accounting errors that nevertheless satisfy the PoolManager’s settlement invariant.

For multiparty computation deployments, Trail of Bits says combining threshold signature schemes with trusted execution environments, or TEEs, can improve security. It cautions that a TEE should provide an additional defensive layer, not replace a sound protocol.

Authentication and infrastructure risks

Trail of Bits argues that SAML, the XML-based authentication protocol that gave rise to the single sign-on industry, has fundamental design weaknesses. It points to XML complexity, canonicalization problems, enveloped signatures and resistance to design changes as reasons signature-wrapping attacks and parser differentials remain possible despite decades of awareness.

The firm recommends migrating to OpenID Connect, or OIDC, citing its simpler JSON-based design, detached signatures and ability to evolve more readily.

Other entries address verification and infrastructure security:

  • Signal: Its recently launched Automatic Key Verification feature helps users confirm chat security without directly comparing safety numbers. Trail of Bits built and operates one of the system’s three auditors.
  • AWS: A post examines security risks associated with integrating AWS Nitro Enclaves and AWS KMS.
  • Package managers: Testing of 25 of the most popular package managers examined whether they could detect a compromised repository inserting malicious code into legitimate packages. Only two checked the integrity of every downloaded package.

AI-assisted patching and open-source maintenance

Patch the Planet is a joint initiative between Trail of Bits and OpenAI intended to help critical open-source software projects respond to vulnerability research conducted at machine speed.

Trail of Bits says its use of Codex is cutting backporting time from days to hours per fix, enabling human patch reviewers to submit critical software fixes more quickly. It is also publishing two plugins, `patching-bugs` and `regression-hunt`, through which engineers can delegate patch development and regression checks to agents.

Drawing on the initiative’s first few weeks, the firm describes three techniques for using `/goal`. The feature gives Codex an open-ended objective and allows it to work independently toward a defined success condition.

Trail of Bits also challenges 1Password’s benchmark, saying it misrepresents AI patching. In that entry, the firm presents real-world evidence on human and agent patch quality from its consulting work and Patch the Planet, alongside the release of two agent skills for testing and reviewing security fixes.

AI research capabilities and containment

A Patch the Planet field report says GPT-5.5-Cyber created a custom zlib fuzzing campaign in one day—a task Trail of Bits says previously required weeks of work by skilled researchers.

Another entry describes how six months of work with AI agents helped uncover real issues during the firm’s Miden zkVM audit. Before code review began, agents helped build custom tools and formal models.

The blog also raises concerns about agent security. Trail of Bits warns that a virtual machine alone can no longer be assumed to contain a sufficiently advanced AI agent. Separately, the firm reports bypassing ClawHub’s malicious skill detector, Cisco’s agent skill scanner and all three scanners integrated into skills.sh.

Cryptographic tools and limitations

Trail of Bits is releasing SequenceHash and the related SequenceMAC function. The two constructions are intended to provide secure multihashing for developers using hash functions other than Keccak.

The firm has also added post-quantum cryptography support to `pyca/cryptography`, described as PyPI’s eleventh most-downloaded Python package. The addition makes the quantum-resistant ML-KEM and ML-DSA algorithms available across the Python ecosystem through a `pip install`.

Other cryptography-related entries report that:

  • A new polynomial-based cryptanalytic technique enabled the rapid factoring of hundreds of weak RSA and DSA keys with biased bits.
  • Cryptographic methods can protect ballot secrecy and help prevent manipulation in electronic voting, but paper ballots remain essential.

Testing tools and formal verification

Trail of Bits is extending Mewt, the open-source mutation-testing engine it released in April, with support for DAML, the language used to write Canton Network applications. Mewt is designed to identify gaps in test suites.

The firm has also added a comprehensive chapter on security testing Rust programs to its Testing Handbook.

For GitHub Actions security, Trail of Bits improved YAML anchor support in zizmor and tested it against 41,253 real-world workflows from high-value open-source projects.

A separate post describes a Lean bug, framing it with a reference to Fermat’s claim that his margin was too small to contain his proof—and suggesting that the bug would have allowed a proof to fit.

TopicSecurity
ProvenanceLive web scan, matched against the project registry
Scope
Canton Network

Sources & attribution

Published
Nov 6, 2026 · 1:23 AM UTC
Source
blog.trailofbits.com

Full attribution under the Digital Millennium Copyright Act (DMCA), 17 U.S.C. § 512: this article is an original report by Canton Intelligence that summarizes the reporting of the publishers named above, links to their original work, and reproduces no protected expression beyond brief factual reference and in-line attribution (“according to blog.trailofbits.com”). All rights in the underlying reporting remain with the publishers listed; nothing here implies endorsement by or affiliation with them. Rights holders may request correction, update or removal through the contact address published at infiniti.w3ai.io.