Services · Security audit
We audit the stacks we spend our days breaking — operating systems, browsers, agentic infrastructure, and the web infrastructure that ties them together. Every engagement pairs senior researchers with Vega, our AI vulnerability research pipeline, so coverage scales without losing depth.
What we audit
01 Kernel and platform review: memory corruption, race conditions, and privilege-escalation paths in the subsystems where they hide — network stacks, filesystems, drivers, and syscall surfaces.
Track record
897 Linux kernel bugs reported upstream.
Multiple KernelCTF wins.
15-year-old GhostLock — the world first Android 17 root exploit.
02 JavaScript engines and the sandboxes around them: JIT miscompilations, type confusions, renderer-to-browser escapes, and the IPC boundaries in between.
Track record
Chrome VRP Top 20.
Multiple Chrome zero-days and counting — from Maglev miscompiles to a single bug piercing both the renderer and the V8 sandbox.
03 The newest attack surface: MCP servers, tool sandboxes, agent frameworks, and model-facing APIs. We chase prompt-injection-to-code-execution chains, tool-permission bypasses, and data-exfiltration paths.
Track record
Multiple vulnerabilities found in major agentic sandbox and browser.
04 Classic surface, modern depth: authentication and session logic, injection, SSRF, deserialization, and business-logic flaws. White-box first, validated with working proofs of concept.
Track record
Findings across WordPress, nginx, curl, and research platforms — every report ships with a PoC.
05 Source-level smart contract review: reentrancy, access control, price-oracle manipulation, upgradeability pitfalls, and the business-logic flaws that drain protocols.
Track record
The same exploit-first standard, applied to code that moves funds.
06 The execution layer itself: bytecode-level review, compiler-introduced bugs, gas and storage-layout semantics, and the places where source-level intuition and on-chain behavior diverge.
Track record
Bytecode-level findings that source review alone misses.
How we work
The report is the artifact, but the exploit is the evidence. Nothing lands in a deliverable that we haven't proven.
A short call to map targets, threat model, and timeline. Fixed quote, no surprises.
Attack-surface enumeration: entry points, trust boundaries, and the code paths that actually matter.
Senior researchers and Vega work the same codebase — manual review for logic, the AI pipeline for scale, fuzzing where it pays.
Every finding is exploited or reproduced before it is written down. No speculative severity.
Root cause, impact, and fix guidance your engineers can act on — plus a retest window after patches land.
By the numbers
0
bugs found and reported
0
CVEs assigned
0
Chrome zero-days
$0K+
in bug bounties earned