Trail of Bits

Go's fuzzer can't solve path constraints, fuzz typed inputs, or catch data races and goroutine leaks.

gosentry is our fork of the Go toolchain that brings a LibAFL engine in Rust, Nautilus grammar fuzzing, struct-aware mutation, and race/leak detection into `go test -fuzz`.

Same harness, stronger engine. We already used it to disclose 4 bugs in Optimism and Revm. https://blog.trailofbits.com/2026/05/12/go-fuzzing-was-missing-half-the-toolkit.-we-forked-the-toolchain-to-fix-it./

We submitted SequenceHash to C2SP. It's a new cryptography spec that prevents a common class of bugs by safely combining multiple inputs into a single hash with any hash function.

Think TupleHash, generalized so it works on top of SHA-256, BLAKE2, or any underlying hash function. The draft is open for review on C2SP, part of our ongoing contributions to open cryptographic standards. https://c2sp.org/sequencehash

30 readers took our C/C++ challenge. Some solved the Linux warmup, but nobody cracked the Windows driver bug. Even LLM-assisted submissions came up short.

The walkthrough explains both, including the Windows escalation from local DoS to kernel code execution.

Best 10 submissions are still getting swag. If you won, we'll be in contact.
https://blog.trailofbits.com/2026/05/05/c/c-checklist-challenges-solved/

RE: @ostifofficial@fosstodon.org

libVLC powers VLC media player, which has been downloaded more than 6 billion times. Our audit produced structural improvements, not just bug fixes. HTTPS for self-update and build dependencies, three new fuzzing harnesses for URL, CSS, and JSON parsing. More in the report.

"You can write exploits for software that exists in only one configuration that one company has. And you can do it on the fly." —CEO Dan Guido in The Verge on how AI collapses the cost of finding bugs. https://www.theverge.com/ai-artificial-intelligence/915660/mythos-script-kiddies-hackers-attack-cybersecurity-ai

libFuzzer is in maintenance mode. We added LibAFL support to Ruzzy so Ruby devs and security researchers can run their next fuzzing campaign without harness modifications.

Adding LibAFL support to Ruzzy took longer than expected. We took detours in ELF file internals, .init_array DSO sections, SanitizerCoverage interceptors, lazy vs. eager loading, and Ruby C extensions. https://blog.trailofbits.com/2026/04/29/extending-ruzzy-with-libafl/

The fastest way to get a team to adopt AI is to make them put in reps. We run hackathons as a forcing function.

2-3 day sprints, one objective. Last time, we told every engineer to use Claude Code in bypass permissions mode. It's now the default for our org.

The full playbook: https://blog.trailofbits.com/2026/03/31/how-we-made-trail-of-bits-ai-native-so-far/

When Claude reasons about code, it reasons about lists, but the questions that actually matter are graph questions.

We just open-sourced Trailmark, a library that parses source code into a call graph using tree-sitter and rustworkx across 17 languages.

8 Claude skills built on its API. On Ed448, one classified 73% of surviving mutants as equivalent. Flat lists can't see that. https://blog.trailofbits.com/2026/04/23/trailmark-turns-code-into-graphs/

"Human plus LLM is vastly vastly better than either one alone."

Our Blockchain Engineering Director Ben Samuels explains why security auditors aren't going anywhere.

Our sales team's AI Maturity Matrix. Scored from 0-3, defining what AI-enabled work looks like at each level.

Adoption is a ladder. Every team has clear levels, clear expectations, a clear path up, and real consequences for staying stuck.

Every org's matrix should look different. Copy the system, not the specifics.
https://blog.trailofbits.com/2026/03/31/how-we-made-trail-of-bits-ai-native-so-far/

Google used a ZK proof to disclose a quantum breakthrough that cuts the cost of breaking cryptocurrency by 20x without handing attackers the circuit.

The Rust code behind the proof had memory safety bugs. We used this new attack surface to forge a proof that beats Google’s on every metric.

Google patched it within days. Their quantum claims are unaffected. https://blog.trailofbits.com/2026/04/17/we-beat-googles-zero-knowledge-proof-of-quantum-cryptanalysis/

RE: @pypi@fosstodon.org

4 billion downloads a day run through @pypi@fosstodon.org. A missing permission check let any org member invite new owners. One of 14 findings from our second audit.

Our C/C++ code review challenge closes April 17.

The new Testing Handbook chapter covers memory safety, integer errors, type confusion, kernel modules, Windows usermode, and seccomp sandbox escapes through manual code review.

Analyze the vulnerable programs, explain how to exploit them, and submit a writeup. First 10 correct entries win swag.

https://trailofbits.com/c-whats-wrong-challenge/

Most companies use AI to do the same work slightly faster. We call that level one adoption. Companies at level three do fundamentally different work. We're somewhere between level two and level three, and it took us a year to get there.

80-95% of our team use Claude weekly. We have 94 plugins containing 201 skills, 84 specialized agents, 29 commands, 125 scripts, and 414+ reference files encoding domain expertise.

The full six-step playbook we used to get there:
https://blog.trailofbits.com/2026/03/31/how-we-made-trail-of-bits-ai-native-so-far/

We open-sourced the system we built to make Trail of Bits AI-first. A six-step playbook for embedding AI into how your team actually works, not just what tools they have access to.

In 8 weeks we went from 5% to 67% Claude Code co-authorship on merged PRs across 59 contributors and 35 repos, from security engineers to PMs to sales. All through systematized adoption any exec can copy.
https://blog.trailofbits.com/2026/03/31/how-we-made-trail-of-bits-ai-native-so-far/

C and C++ run your OS, your browser, your database, and your critical infrastructure. They're also the easiest languages to get catastrophically wrong.

We wrote down everything a security auditor should check: language-level bug classes, stdlib pitfalls, Linux and Windows issues from usermode to kernel, seccomp sandbox escapes, and ptrace handler race conditions.

One checklist, hundreds of checks. https://appsec.guide/docs/languages/c-cpp/

Thousands of CEOs said AI had no impact on productivity. We use AI to catch 200 bugs/week where we used to find 15, and generate $8M per sales rep.

95% of the company pushed back when we started. At unprompted, Dan Guido explains how our 140-person team went AI-native.
https://www.youtube.com/watch?v=kgwvAyF7qsA

Before its launch, we audited WhatsApp's Private Processing TEEs and found 8 high-severity issues (patched). The enclaves yielded to injected config files, unmeasured ACPI tables, spoofed firmware levels, and stale attestation reports.

TEE security is only as good as the implementation details. Four lessons and the full report: https://blog.trailofbits.com/2026/04/07/what-we-learned-about-tee-security-from-auditing-whatsapps-private-inference/

99.86% of 73K+ MBA expressions, simplified. No previous tool could handle all four types.

CoBRA is a new open-source tool that covers linear, semi-linear, polynomial, and mixed MBA obfuscation through 36 specialized passes managed by a worklist orchestrator.

For linear MBAs (the most common), it evaluates on Boolean inputs to fingerprint structure and reconstruct the simplest equivalent. Outputs verified via Z3 or random-input spot checks. https://blog.trailofbits.com/2026/04/03/simplifying-mba-obfuscation-with-cobra/

We saw high coverage mask a fund-draining vulnerability, and caught it with mutation testing. We built the tools to make this routine.

MuTON and mewt introduce bugs, run tests, and find what coverage misses. MuTON supports TON languages. mewt covers Solidity, Rust, Go, and more.
https://blog.trailofbits.com/2026/04/01/mutation-testing-for-the-agentic-era/

93% recall vs 50% for baseline prompts. Our new dimensional-analysis plugin for Claude Code doesn’t ask the LLM to find bugs. It annotates your codebase with dimensional types, then flags mismatches mechanically. Tested against real audit findings. https://blog.trailofbits.com/2026/03/25/try-our-new-dimensional-analysis-claude-plugin/

Adding token A to token B in a DeFi formula is as meaningless as adding meters to seconds. Different dimensions, meaningless result.

Physicists learn this on day one. Smart contract developers rarely think about it, but the same rules apply to on-chain arithmetic.

During an audit, we caught a function passing decimals where assets were expected. Dimensional analysis spotted it instantly. https://blog.trailofbits.com/2026/03/24/spotting-issues-in-defi-with-dimensional-analysis/

A single bug in an ERC-4337 smart account can be as catastrophic as leaking a private key.

We've audited dozens of smart accounts and found six vulnerability patterns that consistently reappear across codebases.

If you're working with smart accounts, each pattern includes safe code examples so you can reference them for your own implementation: https://blog.trailofbits.com/2026/03/11/six-mistakes-in-erc-4337-smart-accounts/

We open-sourced 10 new Claude Code skills from our internal repository.

Including:
agentic-actions-auditor finds security vulnerabilities in GitHub Actions workflows where attacker-controlled input reaches AI agents running with elevated CI permissions.

let-fate-decide draws Tarot cards using cryptographic randomness when your prompt is too vague for a real plan.

git-cleanup categorizes your accumulated branches and worktrees and walks you through safe deletion with gated confirmation.

https://github.com/trailofbits/skills/

We're sponsoring RE//verse in Orlando this week. Sam Sharps, Kyle Elliott, and Julius Alexandre will be there. Come find them if you want to talk reverse engineering or binary analysis. https://re-verse.io/

How do you rebuild a security consultancy around AI without breaking what works? Our CEO, Dan Guido, talks systems, feedback loops, and what it actually takes to go AI-native at [un]prompted on March 4th at 9:10 AM https://unpromptedcon.org/

What if the compiler itself flagged your bugs? Blockchain Engineer, Kevin Valerio, is in Tokyo for SECCON 14 to show how Go’s IR can be modified to catch deterministic bug classes.
If you're attending, Kevin will present from 14:20-14:40 (GMT+9) https://www.seccon.jp/14/ep260228.html

New tool release! Linux memory forensics requires external debug symbols that precisely match your kernel version, symbols rarely installed on production systems and often missing after updates.mquire eliminates this dependency entirely by extracting BTF type information and Kallsyms symbol addresses directly from the memory dump. Works on kernel 4.18+ with BTF enabled.
https://blog.trailofbits.com/2026/02/25/mquire-linux-memory-forensics-without-external-dependencies/

Before launch, Perplexity hired us to test the security of Comet, their AI browser assistant. We demonstrated how four prompt injection techniques could extract users' private information from Gmail. https://blog.trailofbits.com/2026/02/20/using-threat-modeling-and-prompt-injection-to-audit-comet/

Today at 12:55 PM MT on the Future Llama stage at ETH Denver, our CEO, Dan Guido, opens the hood on how he made Trail of Bits AI-native.

Carelessness versus craftsmanship in cryptography
Two popular AES libraries (aes-js and pyaes) provide dangerous default IVs that lead to key/IV reuse vulnerabilities affecting thousands of projects. One maintainer dismissed the issue, while strongSwan's maintainer exemplified proper security response by comprehensively fixing the vulnerability in their VPN management tool.
https://blog.trailofbits.com/2026/02/18/carelessness-versus-craftsmanship-in-cryptography/

We're hiring a senior technical recruiter who can own the full hiring lifecycle and build a talent pipeline. You thrive on personal connections with a knack for evaluating technical candidates across engineering and non-engineering roles.

How to run Claude in YOLO mode safely: Use our devcontainer for full file and network isolation. https://github.com/trailofbits/claude-code-devcontainer

Not isolated enough? We're also sharing dropkit, our custom CLI for quickly accessing DigitalOcean droplets for security testing and research tasks https://github.com/trailofbits/dropkit

Nearly all of our 140 employees use Claude Code daily, and most in YOLO mode. Our devcontainer and dropkit are key linchpins for how we make this safer.

Our team had 375+ pull requests merged into 90+ open-source projects in 2025. From the Rust compiler to PyPI Warehouse to Sigstore, these contributions strengthen the infrastructure devs rely on daily.

Key contributions include:
* rekor-monitor is now production-ready with identity monitoring for Rekor v2
* 20+ Clippy lints merged, including implicit_clone improvements
* pyca/cryptography gained a new ASN.1 API
* PyPI Warehouse now supports project archival

https://blog.trailofbits.com/2026/01/30/celebrating-our-2025-open-source-contributions/

Today's software signatures may not survive tomorrow's quantum computers.
Over the past two years, we collaborated with the Sigstore community to build controlled cryptographic agility into the ecosystem with a centralized algorithm registry, configurable restrictions, and Go implementations of post-quantum algorithms LMS and ML-DSA to prove it's future-ready. https://blog.trailofbits.com/2026/01/29/building-cryptographic-agility-into-sigstore/

New Trail of Bits skill: insecure-defaults
Detect insecure default configurations, hardcoded credentials, and fail-open security patterns. https://github.com/trailofbits/skills/tree/main/plugins/insecure-defaults

TEE security breaks down in predictable ways. In our December webinar, we showed exactly where.
Jules Drean from Tinfoil walked through their threat model, covering repositories, hardware configurations, and CVM images. Our security engineers, Paul Bottinelli and Tjaden Hess, dug into vulnerabilities they've found in production TEE deployments.

Watch the full recording: https://watch.getcontrast.io/register/trail-of-bits-top-tee-bugs-you-should-fix-before-your-audit?utm_source=socials

We open-sourced 17 Claude skills!

Think of Claude skills like Neo's uploads. Install a plugin, and Claude gains the capability in seconds. But we weren't satisfied with an AI plugin that vibes its way to an occasional bug. Our CEO and engineers built skills across the spectrum to see how far AI-assisted security can go.
https://github.com/trailofbits/skills

Browser AI agents resurface web security's old mistakes. We exploited lack of isolation to steal data and hijack sessions. These attacks mirror XSS and CSRF.
Our threat model identifies four trust zones with inadequate controls. Data crosses boundaries unexpectedly, enabling attackers to inject prompts and exfiltrate information. We demonstrated exploits from false information to complete account compromise.
https://blog.trailofbits.com/2026/01/13/lack-of-isolation-in-agentic-browsers-resurfaces-old-vulnerabilities/

We're now accepting applications for our 2026 summer internship program!

Trail of Bits is hiring interns across our software assurance, security engineering, and research & development teams. Over the summer, you'll work on real projects that might include conducting security assessments for critical systems, developing open-source tools, and contributing research that advances the field.

Applications are open now through February! Learn more and apply here!
https://apply.workable.com/trailofbits/j/0C784B6D41/