KNOXLAB

Independent offensive security research

Security research for systems that are difficult to reason about.

I test web applications, APIs, and AI-enabled products, with a focus on authorization boundaries, business logic, and unsafe agent behavior.

Discuss a target See the research focus

Manual testing

Targeted automation

Reproducible evidence

01 / Focus

Where I work

A deliberately narrow practice centered on complex trust boundaries—not a checklist of every security discipline.

01

Web applications
& APIs

Authorization failures, workflow abuse, state transitions, and multi-step logic flaws that scanners and isolated tests tend to miss.

02

AI-enabled
systems

Agent and tool boundaries, prompt injection paths, untrusted context, data exposure, and the application behavior surrounding the model.

03

Research
engineering

Purpose-built recon, source-assisted review, and small automation systems that expand coverage without replacing human judgment.

02 / Method

A finding is useful only when another person can act on it.

  1. 01

    Model the system

    Map actors, assets, state, and trust boundaries before sending payloads.

  2. 02

    Break assumptions

    Test how components behave when identity, sequence, and context stop matching expectations.

  3. 03

    Prove impact

    Reduce the issue to a safe, repeatable path with clear technical and business consequences.

  4. 04

    Hand over clearly

    Deliver evidence, reproduction steps, affected boundaries, and practical remediation context.

03 / About

Independent, manual-first, and evidence-led.

I’m Knox, an independent offensive security researcher working across web applications, APIs, and AI systems. I’m interested in the failures that appear between components: identity assumptions, hidden state, unsafe delegation, and business rules that do not survive adversarial use.

Automation is used to widen coverage and make results repeatable. The actual research remains driven by system understanding, careful testing, and clear evidence.

Available for select private programs, focused assessments, and research collaboration.

04 / Contact

Have a system worth testing?

Send the target context, desired outcome, and timeline. Sensitive details can follow over an agreed secure channel.

cve@knoxlab.co