Deep Learning Security

Curriculum for my Deep Learning Security course offered in Spring of 2020-2023

This course covers the intersection of cybersecurity and deep learning technologies such as CNNs, LSTMs, and GANs. Topics include the application of deep learning to traffic analysis, deepfake detection, malware classification, fooling deep learning classifiers with adversarial examples, network attack prediction and modeling, poisoning attacks, and privacy attacks like model inversion and membership inference. Students will present research papers, perform several exercises to apply attack and defense techniques, and do a final research project. Prior experience with machine learning is required, but necessary details on deep learning will be covered.

Most recent syllabus (2023)

Lecture Slides

• Intro

• Neural Network Basics

• CNNs and Website Fingerprinting

• RNNs and Sequential Security Data

• Representations and Triplet Fingerprinting

• GANs and Deepfakes

• Adversarial ML: Evasion Attacks

• Adversarial ML: Poisoning Attacks

• Attention and Transformers

• Outro

Sample Assignments

• Reading Responses: 1 (evaluation), 2 (related work), and 3 (threat model)

• Website Fingerprinting and CNNs

• Adversarial Examples

• Class Project: Lit Review, Proposal, Final Project, Proposal Presentation, Final Presentation

• Paper Presentation