Sign in

Galois

@galoisinc.bsky.social
54 followers 5 following 114 posts

galois.com For nearly 25 years, Galois has combined mathematical rigor with curious creativity to cut through complexity and guarantee trustworthiness in the world’s most critical systems.

PostsRepliesMedia
Galois @galoisinc.bsky.social · 05/10/2026
Formal methods are becoming an increasingly obvious, powerful choice for defining system boundaries, relationships, and requirements, letting AI fill in the details inside those safe parameters.
111
Galois @galoisinc.bsky.social · 30/09/2026
Foundries may try to steal the design IP of the integrated circuits they are contracted to fabricate. Researchers have tried to thwart this by developing cryptographic mechanisms that hide the original circuit’s functionality. Our latest research reveals a flaw in how these defenses are evaluated.
110
Galois @galoisinc.bsky.social · 30/09/2026
For decades, the hardware security community has relied on cryptographic design-hiding mechanisms to prevent untrusted foundries from stealing integrated circuit (IC) IP. Our latest research reveals a fundamental flaw in how these defenses are evaluated.
210
Galois @galoisinc.bsky.social · 28/09/2026
We’re excited to share the newly released strategic guidance report, “Prevent, Contain, Prove: A Framework for Urgently Securing Critical Digital Systems,” prepared under the auspices of the NDIST Strategic Guidance Committee, co-chaired by Galois Principal Scientist Brad Martin.
120
Galois @galoisinc.bsky.social · 21/09/2026
What if you could change your voice in real time—while mathematically protecting your identity? That’s the idea behind VOXLET, a new privacy-preserving voice transformation technology from Galois and partners. Hear it in action 👇 How it works: www.galois.com/articles/vox...
031
Galois @galoisinc.bsky.social · 17/09/2026
For over 25 years, Galois has advanced the state of the art to solve critical problems for clients including NASA, DARPA, Amazon, NIH, the U.S. Air Force, and Apple.
100
Galois @galoisinc.bsky.social · 14/09/2026
This summer, our fantastic intern Henry experimented with using AI to help translate high-level security properties into formalized properties in 5STARS, our automated network verification tool. Read about his process and takeaways in our latest blog: www.galois.com/articles/usi...
010
Galois @galoisinc.bsky.social · 10/09/2026
Galois is excited to announce the kickoff of HealthyDocs, a groundbreaking project funded by ARPA-H. The growth of the fragmented healthcare data ecosystem has significantly outpaced our ability to secure it. Healthydocs offers a secure, scalable solution. Read more: www.galois.com/articles/ann...
000
Galois @galoisinc.bsky.social · 01/09/2026
In the tech world – and especially for CS students – the pressure to "move fast and break things" using AI is stronger than ever. But what happens when you’re building systems where correctness actually matters?
110
Galois @galoisinc.bsky.social · 25/08/2026
Today’s AI models are optimized to produce useful outputs, not to prove those outputs are correct. Rather than guaranteeing correctness up front, AI relies on post-hoc testing, human review, or waiting for post-deployment failures to catch errors.
110
Galois @galoisinc.bsky.social · 20/08/2026
Signal uses the open source Signal protocol to power end-to-end encryption that keep conversations secure and private. Their official specs explain what this protocol does, but mapping those docs to the actual Rust code in libsignal is tough.
120
Galois @galoisinc.bsky.social · 12/08/2026
This year, Galois hosted our first ever Hiring Our Heroes fellow, Craig Troop, who brought deep experience, curiosity, and a fresh perspective from his 20-year military career. Read on to learn how he fine-tuned AI models to write better Cryptol: www.galois.com/articles/adv...
000
Galois @galoisinc.bsky.social · 05/08/2026
The world of privacy-preserving computation is evolving quickly, and so is Swanky. Learn more in our latest update on Galois’s open-source Rust library suite for secure computation: www.galois.com/articles/swa...
010
Galois @galoisinc.bsky.social · 28/07/2026
Exciting news! Galwegian Sourya Dey and his co-authors won the Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies at PETS 2026! 🏆 Their paper evaluates privacy & re-identification risks in the 2010 U.S. Census. Read the paper: 🔗 hdsr.mitpress.mit.edu/pub/ntchx9im...
hdsr.mitpress.mit.edu
A Simulated Reconstruction and Reidentification Attack on the 2010 U.S. Census
000
Galois @galoisinc.bsky.social · 09/06/2026
Measurement is the foundation of the coming formal verification revolution – building benchmarks that capture the realities of verification engineering, expose where models fail, and drive progress toward industrial-scale verification. Read our latest to learn more: www.galois.com/articles/for...
030
Galois @galoisinc.bsky.social · 03/06/2026
Santiago Cuellar joined Galois as a Research Engineer in 2020 and became a Principal Scientist just five years later. Read his story and learn how curiosity, ownership, and a passion for hard problems helped shape his journey. www.galois.com/articles/fro...
000
Galois @galoisinc.bsky.social · 22/05/2026
We’re excited to announce that SAW now supports generating Isabelle theories from Cryptol specifications – a new capability developed by Galois in collaboration with Apple. www.galois.com/articles/ann...
110
Galois @galoisinc.bsky.social · 13/05/2026
AI is collapsing the barriers to sophisticated cyberattacks – making exploits faster, cheaper, and massively scalable. The solution is not finding bugs faster, but using Formal Verification to mathematically prove that critical software is correct and secure.
120
Galois @galoisinc.bsky.social · 05/05/2026
Big News for the MBSE World: the CAMET® Library is now FREE! As of January 1, Galois changed our CAMET® Library subscription model to be free for up to 10 users, per organization. There is also an Enterprise option for organizations that need more than 10 licenses.
100
Galois @galoisinc.bsky.social · 14/04/2026
This Thursday, April 16, Galois Principal Scientist @m-dodds.bsky.social will be delivering a lecture entitled “What Happens to Software When Proof is Cheap?” as part the Allen School’s Distinguished Lecture Series at the University of Washington.
cs.washington.edu
Distinguished Lecture Series
The Distinguished Lecture Series features prominent scientists and technology figures discussing future research directions.
110
Galois @galoisinc.bsky.social · 02/04/2026
Next Tuesday, April 7, Galois Principal Scientist Katrina Schleisman and co-author Michael Levin will be presenting their paper "On the Conflation of Consciousness and Cognitive Complexity" at the AAAI Spring Symposium on Machine Consciousness in San Francisco. cimc.ai/events/aaai-...
cimc.ai
010
Galois @galoisinc.bsky.social · 31/03/2026
MBSE just leveled up. The AADL Library for SysML v2 brings the best of both worlds together – SysML’s descriptive capacity and AADL’s analysis – within the SysML v2 ecosystem. Learn more: www.galois.com/articles/gal...
Image of a bullhorn, indicating that this is a news release
000
Galois @galoisinc.bsky.social · 25/03/2026
Galois recently wrapped a successful hackathon for the @arpa-h.bsky.social UPGRADE program. The event brought us one step closer towards the program's goal: automated vulnerability discovery, mitigation, and synthesis across complex hospital networks. Learn more: www.galois.com/articles/gal...
011
Galois @galoisinc.bsky.social · 16/03/2026
ZK proofs are headed for real-world deployment — but writing correct ZK statements is tricky. Our new post introduces zkLean, a Lean library that defines a domain specific language for specifying and formally verifying ZK statements. Read more: www.galois.com/articles/zkl...
galois.com
zkLean: A DSL for ZK statement verification
021
Galois @galoisinc.bsky.social · 06/02/2026
We released updated versions of SAW and Cryptol! Learn more: www.galois.com/articles/gal...
galois.com
Galois Releases SAW 1.5 and Cryptol 3.5.0
040
Galois @galoisinc.bsky.social · 02/02/2026
When one of Galois’s most creative engineers retired, two colleagues decided a card wouldn’t cut it — so they built him a steel puzzle safe disguised as a Minecraft gold ore block. Multiple lock layers. Accelerometer-based sequences. Light sensor backup. ~30 lbs of welded steel.
smithdtyler.com
How to Build a Steel Puzzle Safe
Or, three different ways to destroy an Arduino. My friend Ryan and I recently assembled a steel puzzle box styled to look like a Minecraft b...
100
Galois @galoisinc.bsky.social · 14/01/2026
As 2026 begins, we’re proud to reflect on a year at Galois defined by innovation, rigor, and real-world impact. Give our annual "Year in Review" blog a read to learn more about what our scientists have been up to: www.galois.com/articles/202...
galois.com
2025 Year in Review
020
Galois @galoisinc.bsky.social · 08/01/2026
Great to see formal methods, Cryptol, and SAW getting a shoutout from Apple at the HEXACON 2025 keynote! www.youtube.com/watch?v=Du8B...
youtube.com
HEXACON 2025 - Keynote by Ivan Krstić
YouTube video by Hexacon
000
Galois @galoisinc.bsky.social · 18/12/2025
Galois Principal Scientist @m-dodds.bsky.social recently gave a talk at the @toposinstitute.bsky.social on a question we think about a lot: “What actually works (and what doesn’t) when selling formal methods?” Full Talk: www.youtube.com/watch?v=Z2bT...
youtube.com
[Berkeley Seminar] Mike Dodds (Galois) | What works and doesn't selling formal methods in industry
YouTube video by Topos Institute
020
Galois @galoisinc.bsky.social · 26/11/2025
Galois has released updated versions of our Software Analysis Workbench (SAW) and Cryptol, including a wide variety of new features, bugfixes, and enhancements. Learn more: www.galois.com/articles/gal...
Bullhorn icon - indicating news
000
Galois @galoisinc.bsky.social · 17/11/2025
Congrats to Galois spinout Free & Fair, which last week released VoteSecure, a new cryptographic protocol and SDK supporting end-to-end verifiable remote/internet voting. This great example of Rigorous Digital Engineering in action has been featured in several news outlets.
100
Galois @galoisinc.bsky.social · 13/11/2025
At Galois, formal methods form the backbone of everything we do. But what exactly are formal methods? How do they work, and why are they so important? We're so glad you asked. Learn more here: www.galois.com/what-are-for...
000
Galois @galoisinc.bsky.social · 29/10/2025
Last month, Galois Principal Scientist @m-dodds.bsky.social wrote an article about his work exploring using Claude for interactive theorem proving (www.galois.com/articles/cla...). His experiments have continued, with some fun results!
galois.com
Claude Can (Sometimes) Prove It
110
Galois @galoisinc.bsky.social · 27/10/2025
Galois Research Engineer Sourya Dey’s paper “Token embeddings violate the manifold hypothesis” has been accepted to NeurIPS 2025! Come see Dey, with co-authors Michael Robinson and Tony Chiang, in San Diego this December, or dive into the paper here: arxiv.org/abs/2504.01002
arxiv.org
Token embeddings violate the manifold hypothesis
A full understanding of the behavior of a large language model (LLM) requires our grasp of its input token space. If this space differs from our assumptions, our comprehension of and conclusions about...
000
Galois @galoisinc.bsky.social · 24/10/2025
@aria-research.bsky.social has awarded seed funding to Galois for FV-Spec – a new large-scale benchmark of real-world challenge problems to measure and accelerate AI capabilities in formally verifying software. Learn more: www.aria.org.uk/opportunity-...
aria.org.uk
Mathematics for Safe AI
We don’t yet have known technical solutions to ensure that powerful AI systems interact as intended with real-world systems and populations. A combination of scientific world-models and mathematical p...
120
Galois @galoisinc.bsky.social · 13/10/2025
For chipmakers, design-hiding is a $100B problem. Most outsource fab to untrusted foundries—but foundries can fabricate a chip from encrypted designs. In our latest, Tom Shrimpton defines the problem, explores potential solutions, and looks ahead. 🔗 www.galois.com/articles/rev...
galois.com
Reverse Engineer THIS! Towards Provable Cryptographic Protection for Circuit IP
010
Galois @galoisinc.bsky.social · 08/10/2025
With GAI4RDE, Galois developed RDE Wingman: an LLM-based, multi-agent framework that automates RDE workflows. “In a single day, using RDE Wingman could save me 2 weeks of work,” said Galois Principal Scientist Joe Kiniry. “It’s making RDE 10 to 100X faster.” www.galois.com/articles/gen...
000
Galois @galoisinc.bsky.social · 06/10/2025
We’re excited to kick off stage 2 of our project to verify the Jolt zkVM! Thank you to the Ethereum Foundation Ecosystem Support Program for the grant to fund this effort. Check out our post from January to learn more about the project: www.galois.com/articles/tow...
galois.com
Towards a verified Jolt zkVM
011
Galois @galoisinc.bsky.social · 29/09/2025
This summer, as part of a project verifying the frontend of the Jolt zkVM, Galois intern Liza Pertseva and her team used an SMT solver to automate Lean proofs, saving over 6,800 lines of Lean code that would have been otherwise written by hand. Learn more: www.galois.com/articles/aut...
011
Galois @galoisinc.bsky.social · 23/09/2025
This summer, our returning intern – Cornell PhD student Mark Moeller – helped show how Galois’s 5STARS tool could prevent a simulated hospital cyberattack before it ever began. Read more: www.galois.com/articles/gal...
galois.com
Galois Internship Round 2: A Second Summer Intern Experience
000
Galois @galoisinc.bsky.social · 18/09/2025
Once someone understands what formal methods are, their next question is typically "Where do I put them?" Our latest explainer digs in: www.galois.com/where-do-i-p...
100
Galois @galoisinc.bsky.social · 16/09/2025
Interactive Theorem Proving is time-consuming, tedious, and just plain hard. So, Galois Principal Scientist @m-dodds.bsky.social tried using Claude Code for ITP. Surprisingly, IT WORKED (with some big caveats)! Give the full article a read to learn more: www.galois.com/articles/cla...
011
Galois @galoisinc.bsky.social · 11/09/2025
"It’s the people that have made my time here so special," said Galois intern Angel Gallegos. "Galois is full of really smart people, but everyone is down-to-earth and happy to talk. You’re always learning something new." Read more in our Q&A with Angel: www.galois.com/articles/gal...
000
Galois @galoisinc.bsky.social · 02/09/2025
Big LLMs are great... unless you're working with sensitive data. Galois intern Kevin Fisher spent the summer exploring how far local LLMs can go in code generation without sacrificing privacy. Read Kevin's full writeup here 👇 www.galois.com/articles/pri...
000
Galois @galoisinc.bsky.social · 25/08/2025
Big news: Brad Martin has joined Galois as our newest principal scientist! After nearly 40 years in federal service, Brad is bringing his vision + expertise in formal methods & cybersecurity to help us tackle some of the world’s biggest tech challenges. Welcome to the team, Brad!
000
Galois @galoisinc.bsky.social · 18/08/2025
This summer, Galois intern @binarynewts.bsky.social explored how LLMs could help translate legacy C code into safe Rust. Their project, CNnotator, generates memory safety annotations for C code. Learn more: galois.com/articles/esc...
031
Galois @galoisinc.bsky.social · 14/08/2025
Evaluating FHE program performance used to take weeks of trial-and-error. Dioptra (open-source from Galois) streamlines this process—estimating time & memory usage for OpenFHE Python programs, with performance models & comparisons to guide better decisions. 🔗 www.galois.com/articles/gal...
galois.com
Galois Releases Dioptra: An analytic tool for OpenFHE programs
000
Galois @galoisinc.bsky.social · 07/08/2025
"Galois is unlike any place I’ve worked before," writes Galois intern Mariana Ferreira. "There is a remarkable level of humility across the organization. Titles and hierarchies take a backseat to curiosity, kindness, and collaboration." Read more: www.galois.com/articles/ref...
galois.com
Building the Future, Securely: Reflections on My Summer at Galois
000
Galois @galoisinc.bsky.social · 24/07/2025
Congrats to Galwegians Scott Moore, David Darais & Sourya Dey (and former Galwegians Ethan Lew & Ramy Tadros) on “A Simulated Reconstruction and Reidentification Attack on the 2010 U.S. Census” being accepted at Harvard Data Science Review. Read more: hdsr.mitpress.mit.edu/pub/ntchx9im...
hdsr.mitpress.mit.edu
A Simulated Reconstruction and Reidentification Attack on the 2010 U.S. Census
Forthcoming. Now Available: Just Accepted Version.
110
Galois @galoisinc.bsky.social · 22/07/2025
🚨 Paper accepted! Congrats to Galwegians Santiago Cuéllar, James Parker, Stuart Pernsteiner + co-authors Bill Harris, Ian Sweet & Eran Tromer on "Cheesecloth: ZK Proofs of Real-World Vulnerabilities” being accepted for publication in ACM TOPS! Read it!: dl.acm.org/doi/10.1145/...
dl.acm.org
Cheesecloth: Zero-Knowledge Proofs of Real-World Vulnerabilities | ACM Transactions on Privacy and Security
Currently, when a security analyst discovers a vulnerability in critical software system, they must navigate a fraught dilemma: immediately disclosing the vulnerability to the public could harm the sy...
010