Shipped: Adversarial AI — Defensive Models and Robustness Experiments

An open-source collection of defensive AI research artifacts supporting Book 3 of The Applied AI Universe Coding Guide series.

The project examines how machine learning systems can be protected against adversarial attacks and how those defenses perform when attackers adapt.

The associated research framework covers 35 classical and hybrid/quantum victims, spanning machine learning, deep neural networks, reinforcement learning, generative AI, natural language processing, and quantum computing.

Defensive approaches include adversarial training, differential privacy, data sanitization, anomaly detection, model integrity verification, provenance checking, input normalization, and checkpoint recovery.

The experiments reconstruct attacks from Book 2, apply corresponding defenses, measure performance, and then reevaluate each system against an adaptive adversary who knows the defense is present.

Importantly, the collection preserves unsuccessful and partially effective defenses to demonstrate the limitations of individual security controls.

Designed for AI security researchers, educators, and developers studying reproducible adversarial robustness experiments. These are educational research artifacts, not certified production defenses or evidence of a general quantum advantage.