For teams building parsers, libraries, SDKs, protocols or application components that process untrusted input.
When malformed files, messages or inputs could expose a weakness.
What we examine
- Assess target feasibility, build requirements, entry points and test objectives.
- Set up and run the agreed testing approach for the selected component.
- Investigate and deduplicate failures, minimize useful examples and assess security relevance.
How an engagement runs
Send a short brief
The product, what changed, what you need to know and by when. A paragraph is enough, and I reply within three days.
Get a written proposal
Scope, method, exclusions, deliverables, price and dates. Confidentiality, NDA and retesting terms are agreed in the same proposal.
Follow the work as it develops
For assessments, you see evidence and priorities as they emerge. For engineering, you review the agreed milestones.
Report and walkthrough
You receive the report or the delivered work and a walkthrough with your team. Any further fixes or verification are agreed with you.
I find the cause, then write the fix.
I wrote the patches behind two Linux kernel CVEs and seven more fixes now in mainline. On your product, each finding comes with the code path, the conditions that trigger it and a fix your team can review.
See the 9 merged kernel patchesQuestions before we begin
What makes a useful fuzzing target?
A component with clear input boundaries and a build or execution environment we can exercise. Parsers, libraries and protocol handlers are useful candidates. We confirm suitability before committing to a campaign.
Can we start with a small pilot?
Yes. A paid feasibility phase can establish whether the target is practical, what setup it needs and what a larger campaign should include.
Will you deliver a harness or ongoing test setup?
We agree the artifacts during scoping. Depending on the target, these may include a harness, seed inputs, run instructions, reproducers or integration guidance. They are specified in the proposal.
Does every crash mean a vulnerability?
No. Failures need investigation. The report distinguishes observed behavior, confirmed security impact and unresolved questions; it does not promise a number of vulnerabilities or CVEs.
How quickly will you reply?
Within three days at most. You get questions about your product or an outline of the proposal.
What does it cost?
Published prices are starting points. The final price depends on scope and technical complexity and is fixed in the written proposal before work starts.
How long does it take, and when can you start?
Duration and start date depend on the scope, the technical complexity and my current workload. The proposal states the dates.
Can we work under an NDA?
Yes. Confidentiality terms, including an NDA if you need one, are agreed as part of the proposal before you share code, credentials or findings.
Is a retest included?
Retesting is agreed for each engagement in the proposal: which findings, which version and the time window.
Do you work with agencies?
Yes. Agencies can bring me in for a client’s project. Delivery terms, including white-label reports, are agreed case by case in the proposal.
Related reading
- CVE-2026-98432: one function swapped for another, twice, and why disabling MPTCP isn’t the fix
- PHP2Shell: CVE-2026-109618 and the Runtime Boundary
- Lovable and Supabase security checklist: what to check before launch