CIS 482/582: Trustworthy Artificial Intelligence

University of Michigan, Dearborn

Schedule and Materials

Take the below schedule as tentative, depending on progress it will be updated as the semester advances.

Week Topic Slides/Demos Resources/Suggested Reading
1 Motivation and Intro Slides
Video
  1. Birhanu Eshete, Making Machine Learning Trustworthy
  2. Kush R. Varshney, Trustworthy Machine Learning and Artificial Intelligence
2 A Crash Course on Deep Neural Networks Slides
Video
Demo
  1. François Fleuret. The Little Book of Deep Learning
Machine Learning Attack Surface No separate lecture for this: it is covered within adversarial examples, training data poisoning, membership inference, and model stealing
  1. Papernot et al., SoK: Security and Privacy in Machine Learning
  2. Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations
3 Adversarial Examples Slides
Video
Demo
  1. Szegedy et al., Intriguing properties of neural networks
  2. Papernot et al., Practical Black-Box Attacks against Machine Learning
  3. Eykholt et al., Robust Physical-World Attacks on Deep Learning Visual Classification
  4. Goodfellow et al., Explaining and Harnessing Adversarial Examples
  5. Amich and Eshete, Morphence: Moving Target Defense Against Adversarial Examples
4 Training Data Poisoning Slides
Video
Demo
  1. Gu et al., BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
  2. Chen et al., Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
  3. Shan et al., Poison Forensics: Traceback of Data Poisoning Attacks in Neural Networks
5 Membership Inference Slides
Video
Demo
  1. Shokri et al., Membership Inference Attacks against Machine Learning Models
  2. Papernot et al., Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data
  3. Abadi et al., Deep Learning with Differential Privacy
  4. Jarin and Eshete, MIAShield: Defending Membership Inference Attacks via Preemptive Exclusion of Members
6 Model Extraction Slides
Video
  1. Tramer et al., Stealing Machine Learning Models via Prediction APIs
  2. Ali and Eshete,Best-Effort Adversarial Approximation of Black-Box Malware Classifier
  3. Jia et al., Entangled Watermarks as a Defense against Model Extraction
7 Transparency and Interpretability Slides
Video
Demo
  1. Rudin Cynthia. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead
  2. Ribeiro et al., “Why Should I Trust You?” Explaining the Predictions of Any Classifier
  3. Scott Lundberg, Su-In Lee. A Unified Approach to Interpreting Model Predictions
8 Fairness Slides
Video
Demo
  1. Dwork et al., Fairness Through Awareness
  2. Zemel et al., Learning Fair Representations
  3. Hardt et al., Equality of Opportunity in Supervised Learning
  4. Buolamwini and Gebru, Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification
9 Ethics and Governance Slides
Video
  1. NIST: AI Risk Management Framework (AI RMF 1.0)
  2. Weidinger et. al., Taxonomy of Risks posed by Language Models
10 Holistic Trustworthiness Considerations and Open Issues Slides
Similar Courses: Below are similar courses on the topic of trustworty AI/ML. Depending on who teaches a course and the institution, depth and breadth of topics may vary.

© Birhanu Eshete 2024