Independent offensive security research

Offensive security research at the seams.

ml supply chain · inference runtimes · automation platforms

I'm Nathan, a penetration tester and researcher. Here you'll find projects, interesting links and notebook entries based on current research initiatives.

Notebook

Notes from the bench

2026-05-01 · ml-security

Channel, Decoder, Substrate: A Vocabulary for ML Attacks

An ML attack is the composition of three things: a channel that carries information, a decoder that reads the channel, and a substrate that runs the decoder. Naming the shape changes how you think about both offense and defense.

Open notebook →

Projects

Active research · Tools · Papers in progress

Manta Weight steganography: covert channels in neural network weights and the defenses against them. Paper-staging Crucible The first structure-aware fuzzer for ML model formats and inference infrastructure. Shipping Scorch Offensive toolkit for Microsoft System Center Orchestrator: the credential vault nobody had tooled up. v2.0.1 Hemlock Weaponize documents. Test RAG defenses. Harden pipelines. Shipping halo-forge RLVR training on consumer AMD silicon. Shipping malagent Exploring RLVR for security research. Research preview All projects →

Signal

Curated reading and monthly digests

Curated reading and monthly digests, populated as I add them. The full Signal page is the working surface; this section will surface latest items once there's something here.

All signal →