←
Oak Security2,203 posts
Oak Security avatar✓

Oak Security ✓

@SecurityOak

Book. Secure. Relax. Oak Security offers audits, penetration testing, training, and advisory.

◷ Joined July 2021▧ 325 Photos and videos
Oak Security avatar
Oak Security@SecurityOak·6h

Keep the comparison honest: published findings and exploit incidents are different datasets. They cannot tell us how many attacks audits prevented. Use the talk to question coverage assumptions and plan the security work around the audit.

Oak Security avatar
Oak Security@SecurityOak·6h

Take one path to funds. Trace it through the contracts, signer devices, approval process and recovery procedure. Mark what your planned review covers. Give each uncovered part an owner and a review method.

Oak Security avatar
Oak Security@SecurityOak·Nov 15

Planning an audit? Tell us what you're building, what is changing and which assumptions worry you. Talk to Oak about human-led, AI-assisted audits and how TRACE can help define the scope. oaksecurity.io/contact

Oak Security avatar
Oak Security@SecurityOak·Nov 14

At Oak, human researchers agree the scope, manually audit the code and verify findings. Hive supports discovery and adversarial validation. Reach out to talk through your next release.

Oak Security avatar
Oak Security@SecurityOak·Nov 14

Add privileged roles, external dependencies and the assumptions you want challenged. Mark anything that changed since the last review.

Oak Security avatar
Oak Security@SecurityOak·Nov 14

Preparing for a human-led, AI-assisted audit? Start with the exact revision and the behaviour your system must preserve.

Oak Security avatar
Oak Security@SecurityOak·Nov 13

A compromised account can become a malicious build, then a production change. TRACE follows authority across protocol rules, systems and people. For each boundary, ask what still holds if the source is compromised. Methodology: github.com/oak-security/TRAC…

Post attachment
Oak Security avatar
Oak Security@SecurityOak·Nov 11

Two models agreeing can still mean two models sharing a blind spot. Hive keeps discovery separate and records the reasons behind validation. A senior researcher reviews the code and the evidence before accepting a finding.

Oak Security avatar
Oak Security@SecurityOak·Nov 10

TRACE makes that review explicit. AiTM applies the workflow with AI assistance; we're preparing its open-source release. The researcher still needs to judge whether the analysis holds.

Oak Security avatar
Oak Security@SecurityOak·Nov 10

Then inspect the path. What must the attacker control? Which invariant would fail? Which step is assumed rather than established?

Oak Security avatar
Oak Security@SecurityOak·Nov 10

What should you ask of an AI-assisted threat model? Start with the evidence: which source supports this actor, permission or trust boundary?

Oak Security avatar
Oak Security@SecurityOak·Nov 9

An architecture diagram shows connections. A threat model also needs to show authority. Who can change the build, approve a transfer or recover an account? Follow those paths across protocol rules, infrastructure and people. That is where TRACE starts.

Oak Security avatar
Oak Security@SecurityOak·Nov 8

We're preparing to open-source AiTM, our AI Threat Modeller, built on TRACE. It uses AI to turn specs and code into a structured threat model for human review. Release details will follow. Register your interest: oaksecurity.io/ai-threat-mod…