Bayesian spam classification: the dataset
Preparing the SMS Spam Collection dataset for Bayesian classification, covering download, extraction, loading, and cleaning through an adversarial lens.
Spam classification: Naive Bayes filters
How Naive Bayes spam filters work, why the independence assumption makes them exploitable, and how GoodWords attacks broke email filtering wide open.
Metrics for evaluating a model
Learn how accuracy, precision, recall, and F1-score work in practice, where each metrics deceive, and how adversaries exploit the gaps they leave behind.
Python libraries for AI red teaming
Python Libraries: How scikit-learn and PyTorch work, and why their APIs are the operational foundation for adversarial machine learning.
Getting started with JupyterLab
Get started with JupyterLab: interactive notebooks for data analysis. Create, visualise, and document your Python work.
Setting up an AI environment on Apple Silicon
Three common problems when setting up PyTorch and Miniconda on M-series Macs for AI, from the conda zsh mismatch to CUDA commands that don't belong on your Mac.
Diffusion models
The latest entry in the AI red teaming series breaks down how diffusion models work and maps five distinct attack surfaces across the generation pipeline.
Large language models (LLMs)
Every stage of an LLMs inference pipeline creates exploitable behaviour. Tokenisation, embeddings, attention, and decoding explained for the adversarial mind.
Generative AI
Generative AI: How GANs, VAEs, autoregressive models, and diffusion models work, and the specific attack surfaces each architecture exposes to AI red teamers.
Recurrent neural networks
How recurrent neural networks process sequences, where their memory breaks down, and what that means for red teaming sequential security systems.