The intersection of security and deep learning covers two primary areas: using deep learning to security (e.g., intrusion detection) and protecting deep learning models from vulnerabilities (e.g., adversarial attacks) . Key Security Threats to Deep Learning
: Subtly altering input data to trick a model into making incorrect predictions.
: Reverse-engineering a trained model to reveal its parameters or architecture.
: Reconstructing sensitive training data from a model's predictions to compromise privacy. Deep Learning for Defense
Researchers focus on several critical vulnerabilities that can compromise AI models:
