Oakland University · Fall 2026

CSI 4580/5580: AI for Cybersecurity and Privacy

Foundations of modern machine learning, adversarial security, privacy-preserving learning, fairness, and trustworthy AI.

Instructor: Dr. Mohamed Seif

Course Objective

This course introduces the principles and applications of artificial intelligence (AI) for cybersecurity and privacy. Beginning with a hands-on introduction to deep neural networks and AI modeling, students develop the foundations needed to understand modern AI systems and their applications in cybersecurity.

Building on these foundations, the course examines the security of AI systems, including adversarial machine learning, data poisoning, model evasion, model extraction, privacy inference attacks, and defenses. It also explores privacy-preserving machine learning, explainability, fairness, and ethical considerations, emphasizing the opportunities, limitations, and tradeoffs involved in deploying trustworthy AI in security-critical settings.

Course Structure

Module 1 — A Crash Course on Modern Machine Learning (ML)

Core ML and deep learning concepts, neural networks, optimization, training, and evaluation.

Module 2 — Robust and Adversarial ML

Adversarial machine learning and techniques for attacks and defenses.

Module 3 — Privacy Preserving Learning

Basics of differential privacy and privacy-preserving machine learning.

Module 4 — Fairness in ML

Notions of fairness and algorithmic techniques for learning fair classifiers.

Module 5 — Emerging Topics in AI

AI-Enabled Cybersecurity and Privacy

Module 6 — Student Project Presentations

Presentation and discussion of semester-long research projects.

Course Schedule

Lecture Topic
1 Syllabus and Course Logistics
2 Optimization in Machine Learning — Part I
3 Optimization in Machine Learning — Part II
4 Adversarial Machine Learning — Part I
5 Adversarial Machine Learning — Part II
6 Data Poisoning Attacks
7 Membership Inference Attacks
8 Model Inversion Attacks
9 Explainable AI and Fairness
10 & 11 Differential Privacy
12 Privacy-Preserving Learning
13 Adversarial Genomic Sequences (Student Lecture)
14a Adversarial ML for Email Spam Detection (Student Lecture)
14b DP-SGD and PATE
15 Agentic Software Development Life Cycle (Student Lecture)
16 Model Inversion Attacks Against Collaborative Inference (Student Lecture)
17 Differentially Private Prompt Perturbation for Multi-turn LLM (Student Lecture)
18 Extracting Training Data from Large Language Models (Student Lecture)
19a Fairness in AI: Fundamental Notions and Definitions
19b Joint Design of Trustworthy AI: Balancing Competing Objectives
20 & 21 Understanding Transformers: From Architecture to Trustworthiness Applications
22 Self-Attention Mechanisms for Cybersecurity Applications
23 AI-Powered Phishing Detection — From Email Text to Multimodal Threat Analysis

Lecture titles reflect the Moodle listing as of November 12, 2026. Check Moodle for lecture files, assignments, and subsequent updates.