Archive

Prompt injection classifiers are terrible security boundaries

What you actually download when you download a model

Malicious Job Assessments

From HATEOAS to MCP: Guiding AI Agents Through State Machines

Some notes on ML-Based Portfolio Management

Differentially Private finetuning for LLMs

Privacy-Preserving Canvas Fingerprinting

Backdooring Linux with Linker Envs the right way

Short story about evading Antivirus Detection

Brief introduction to Differentially Private Machine Learning

3D-GAN