Real bugs I found, written up in full: the recon, the exploit, and how it played out. Mostly web, mobile, and AI. The whole archive is below.
The frontend JavaScript is a map of the backend: every API endpoint, hidden parameter, and sometimes a leaked key. How I collect a web app's JS, pull endpoints and secrets out of it, use source maps, and turn it into undocumented attack surface.
Read writeupVertical and horizontal subdomain recon, end to end: the passive sources, the active brute forcing, why different tools find different results, and why a root domain beats an IP for mapping what a company owns.
Read writeupHow I run an AI security assessment end to end, mapping what the model can reach, where I look for bugs, and what "secure" realistically means for an LLM app.
Read writeupPrompt injection is the bug I find most on AI apps. How I test for it, what indirect injection looks like, and how to tell a real finding from a weird answer.
Read writeupRAG apps answer from your own documents, which is exactly where the bugs hide. Reading documents you shouldn't, and poisoning answers by planting content.
Read writeupAgents can act, not just talk. How I test tool abuse, excessive agency, and the attacker-controlled inputs that make an agent do things its user never asked for.
Read writeupMost AI guardrails are thinner than the team thinks. How I get past them, and how to tell whether a jailbreak actually matters for your app.
Read writeupThe AI bug that isn't about AI: model output dropped into a page, query, or shell turns into XSS, SSRF, or RCE. How I trace and test it.
Read writeupHow much will an AI app hand over, its system prompt, its keys, other users' data? How I pull secrets out, and why the fix is limiting what the model can see.
Read writeupThe risk that's baked in before anyone types: poisoned training data, backdoored models, and untrusted weights. Treating your AI stack as a supply chain.
Read writeupThe short version: the checklist I work from on every LLM app, from blast radius to injection, leakage, output handling, agents, and cost abuse.
Read writeupMapping AI security testing to the frameworks people answer to, OWASP LLM Top 10, NIST AI RMF, and the EU AI Act, and what a pentest does and doesn't cover.
Read writeupSecurity teams face a simple problem: too much data. A look at how AI augments analyst capability, and where it should never replace human judgment.
Read writeupRuntime instrumentation with Frida, deeper attack surface, and the techniques worth knowing for serious iOS assessments.
Read writeupA foundational guide to setting up an iOS pentest lab, toolchain, jailbreak basics, IPA decryption, and the early signal-rich attack surface.
Read writeupAdvanced techniques with Frida & Objection, hooking, bypasses, and patterns for tougher Android targets.
Read writeupStep-by-step introduction to Android application assessment, setting up the lab, fundamental concepts, first targets.
Read writeupHow a misconfigured Cloudflare tunnel exposes the origin, and how that exposure gets weaponized in real engagements.
Read writeupA real file-upload vulnerability from a pentest engagement, interception, mutation, and the full exploitation path.
Read writeupA reflected XSS discovered during a live engagement, escalated all the way to a full account takeover via session theft.
Read writeupA SQL injection in a US government property, the discovery, the responsible disclosure, and what the dump revealed.
Read writeupA blind SSRF on a major brand's asset, the discovery, the validation chain, and a successful Red Bull bug-bounty reward.
Read writeupFirst writeups in this category are in progress — check back soon, or reach out if there's something specific you want to see.
Get in touchWeb, mobile, API, cloud, or your AI/LLM apps. Let's see what's actually there.