Sam Austin AI·Oct 9Adversarial Attacks Explained: FGSM and How to Defend Against ItAdd a barely visible pattern of noise to a photo of a panda, and a well-trained image classifier confidently calls it a gibbon. To a human…
mihiran vikram·Oct 6Adversarial review of autnomous coding agent on cloudI built a system that lets AI coding agents write and ship software on my own cloud, around the clock, with no one watching. Then I had it…
Shreyapanwar·Sep 29Adversarial attacks on deep learning modelsHave you ever wondered what would happen if a machine learning model is purposely fooled with deceptive data just so it would produce wrong…
Grama Tudor-Ionuț·Aug 30Jacobian Saliency Map Attack on Colour ImagesA dog becomes a cat by changing two of 3072 values. Revisiting the Jacobian Saliency Map Attack on colour images.
Deon Binny·Aug 20Adversarial Machine Learning: 40 Attacks, 20 Defenses, and What I Learned About ML SecurityFrom an Elective Subject to a Hands-On Machine Learning Security Project
Ananyakar·Aug 7Watching the User, the Web, and the Model Itself: Biometrics, Web Security, and Adversarial MLPart 1: Behavioral Biometrics — Continuous Authentication
Christian Jose Bernardo Echeverría Ramírez·Jul 31Breaking an Object Detector: A Gradient-Based Adversarial Attack on YOLOv8-OBBHow I made an aerial object detector stop seeing a vehicle — not by covering it, but by optimizing a texture against the model itself.
Devrushishrutika·Jul 24AI Security: Protecting Large Language Models (LLMs) from Prompt Injection and Adversarial Attacksby Shrutika Devrushi
InTowards AIbyMaedeh Torkian·Jul 14Learning to Say “I Don’t Know”: The First Floor of AwarenessEssay #4 in the Humble Model SeriesA response icon1A response icon1
Stoic_Gang·Jul 9How FGSM Actually Works: The Formula, UnpackedOne line of math. Four moving parts. A model that breaks in a single step.
