Solutions

Product

Evidence Lab (private preview)

Two things are changing at once. The move to post-quantum cryptography will replace the cryptography in nearly every system in use. And AI agents now find new ways to attack cryptographic systems much faster than people could alone, so a system that was sound last year may not be sound now.

Defence has to keep the same pace. The research into how a system can be attacked, and the fixes and patches that answer it, need to move as fast as the attacks do, with a person approving every change.

Evidence Lab helps security teams do that with proof at each step, not assumptions.

  • Scan your cryptographic inventory and all of its dependencies, or bring in the inventory you already have, and check which of its claims hold.
  • Test the systems that matter against current attack research, before and after a change.
  • Get a plan for each system, with recommended fixes and patches: keep classical for now, go hybrid, or move fully to post-quantum.
  • Nothing changes until a person approves it.
  • Keep a signed record of what was tested, when, with what result, and who approved it.

Why it has to keep checking

The time between a weakness existing and someone using it is shrinking, for two reasons.

AI is making it cheaper to find weaknesses in how cryptography is built and configured. Google's Big Sleep agent found CVE-2025-6965, a critical flaw in SQLite that Google says was known only to threat actors and at risk of being exploited (report). In the final of DARPA's AI Cyber Challenge in 2025, autonomous systems analysed 54 million lines of open-source code, found 18 real vulnerabilities and patched 11 of them, at an average cost of $152 per task (report). Most cryptographic failures are of this kind: a missing check, an outdated library, a weak setting. They are not breaks of the underlying mathematics, and no public result shows AI breaking a well-designed algorithm.

Quantum computers would attack the mathematics itself. NIST proposes to deprecate RSA and elliptic-curve public-key cryptography after 2030 and to disallow it after 2035 (NIST IR 8547), and traffic recorded today can be decrypted once such a computer exists.

A review done once a year keeps pace with neither. Evidence Lab is built to check continuously. Every result, fix and retest is kept as signed evidence that someone other than us can verify, and nothing changes a system until a person approves it.

Evidence Lab is in private preview. To ask about early access, write to kevin@isogenylabs.io.

Services

Hardness and parameter review
You have a curve, an isogeny parameter set, or a custom group, and you need a security level you can defend. We estimate the cost of the best known attack, and where the numbers are close we run a scaled-down version to check the estimate.
Elliptic-curve and isogeny implementations
Reference implementations and test vectors for SQIsign, CSIDH-family schemes, pairings, and curve arithmetic, tested against an independent model rather than against themselves.
Post-quantum migration
Finding where your systems depend on RSA and elliptic curves, planning a hybrid move to ML-KEM and ML-DSA, and working out how to roll back if interoperability breaks partway.
Protocol and implementation review
The protocol, its security proof if it has one, and the code: nonce generation, key handling, and side channels. Findings are ranked by how exploitable they are.

Each engagement starts with a written scope that names the system, the adversary, and what would count as a break. At the end you get a report and any code we wrote, and the code runs without us.