CyberResearchLab
We hunt what no scanner has a signature for: pre-auth bugs on the internet edge, memory corruption in parsers, logic flaws in identity flows. We prove exploitability end to end, coordinate fixes with vendors, and are researching an AI model that reasons like an exploit developer.
01 Mission
Scanners find what is already known. We look for what isn’t.
Signature-based testing is bounded by the CVE feed. We work from first principles instead: taint from attacker-controlled input to dangerous sinks, binary diffs of vendor patches, snapshot fuzzing under sanitizers and reverse engineering of closed firmware, until a hypothesis becomes a deterministic proof of concept.
Every result runs the same pipeline: vendor first, then a CVE record and a technical advisory, then a detection or exposure check in the ResilientX Security platform. Next, an AI model that runs that pipeline at machine speed, with researchers in the loop.
- 01
Zero-day research
Original vulnerability research in edge appliances, enterprise software, parsers and cloud control planes: the pre-auth attack surface where a single bug is a breach.
Explore - 02
Exploit development
Proving impact instead of guessing it: from a crash to a primitive, from a primitive to code execution, against real mitigations on production configurations.
Explore - 03
Disclosure & CVEs
Vendor first, always: private reports to the vendor’s PSIRT, coordinated embargoes, CVE records through a CNA, and advisories with root cause and detection guidance once a fix ships.
Explore - 04
Threat research
How real attackers operate: n-day weaponisation timelines, exposed infrastructure at internet scale and new exploitation techniques, turned into detections and exposure checks.
Explore
02 AI researchIn research
An AI model that reasons like an exploit developer
We are researching a model specialised for offensive security. It reads decompiled code, writes fuzzing harnesses, triages crashes to their root cause and plans exploitation, and every hypothesis it makes is verified by execution in a debugger, a fuzzer or an emulator before a researcher sees it.
- 01Code and binary reasoning
- 02Harness and grammar synthesis
- 03Crash triage and root cause
- 04Exploit planning
03 Method
From hypothesis to advisory
- 01
Map
Enumerate the attack surface: pre-auth endpoints, IPC and RPC interfaces, file and protocol parsers, and the privilege each trust boundary runs with.
- 02
Hunt
Taint-guided code review, CodeQL and Semgrep variant analysis, binary patch diffing, structure-aware and snapshot fuzzing under sanitizers.
- 03
Prove
Minimise the crash, establish the primitive, and build a deterministic proof of concept against a release build with default mitigations on.
- 04
Disclose
Private report to the vendor’s PSIRT with root cause, PoC and CVSS v4.0 vector; coordinate the fix, the CVE record and the embargo.
- 05
Publish
Full advisory once patched: root cause, affected and fixed versions, detection guidance and the complete disclosure timeline.
04 Responsible disclosure
Vendor first. Always.
Every vulnerability goes privately to the affected vendor’s PSIRT, under an agreed embargo, and is published with a CVE record and a full technical advisory only once users can protect themselves.
How disclosure works05 Engage
Request an
engagement
Free for open-source products
Commission the Labs team for white-box or black-box research on a product, device or service you build or depend on: zero-day hunting, exploitability assessment or threat research. Tell us the target and we will scope the work with you.
Found a vulnerability in our products? Report it privately to security@resilientx.com · security.txt