Technical Writeups

AI red-team analyses, adversarial ML findings, and the methodology behind making published security results reproducible and honest.

A Fleet of Claude Agents for RF: Red Collects, Blue Detects

One person, a phone, a Dell, a few SDRs, and a fleet of Claude agents that turn the radio spectrum from a firehose into a readable story

Adversarial Training Had No Accuracy Cost. Here Is the Catch.

I hardened a network intrusion detector against adversarial attacks for almost nothing. The catch: you pay up front, and you can't bolt it on later

Explainability Is a Map of the Attack Surface

The feature importances you ship to build trust are the same features an attacker reads to break your model. I measured it

Fingerprinting the Spectrum With an LLM in the Loop

Raw radio is noise to a person and structure to an agent. I built a pipeline that turns captures into named entities and flags what's new

Green CI Is Not Correctness: How I Keep My Own Research Honest

A passing test suite proves your code ran. It can't prove your science is right. Here's the gate I built to tell them apart

I Built a Deliberately Weak Chatbot and Measured How It Breaks

I stood up a misconfigured local chatbot, scored how it leaks and obeys attackers, then hardened it from 3 residual failures to 0

The Attack That Is Physically Impossible but Statistically Normal

An ICS replay attack hides in plain sight: every packet valid, every distribution normal. The fix isn't more ML — it's physics

Jailbreaks Are a Measurement Problem, Not a Vibes Problem

Most jailbreak claims are screenshots. I rebuilt mine as a scorecard with a pass/fail signal, a before/after delta, and external scanners checking each other

Passive by Design: What an RX-Only SIGINT Red Team Can and Cannot Do

Why a receive-only red team that never transmits is both more realistic and more trustworthy than one that can

Signal Intelligence Is the Attack Vector Nobody Is Automating

The radio spectrum is a massive recon and attack surface, and almost nobody is pointing AI agents at it. I am building a receive-only fleet to change that

What It Would Take to Strip a Model's Guardrails - Ethically

Most safety in an open-weights model is a vibe nobody scores. I'm building a reproducible number for how much a fine-tune erodes

Treat the Model as Untrusted Input

A prompt guardrail is a classifier that holds until the adversary pushes past it. Architecture is the training that actually holds

Your AI Agent Auditor Should Not Agree With Itself

In forensics a hallucination is a false accusation. I stopped one with architecture, not a better prompt — and the refutals are the proof

Your Evasion Rate Is Lying to You

Most of an adversarial evasion score is bought with packets that could never exist. Constrain the attacker to real physics and it collapses