Lecture 4 Additional Software Tools

• Adversarial Robustness Toolbox
• Counterfit
• CleverHans
• foolbox
• Privacy Meter
• NIST SP 800-188: De-Identifying Government Datasets
• robustness
• TensorFlow Privacy

Lecture 4 Additional Software Examples

• Attacking a Machine Learning Model
• Machine Learning for High-Risk Applications: Chapter 11 code examples (red-teaming XGBoost)

Lecture 4 Additional Reading

• Introduction and Background:

  • A Marauder’s Map of Security and Privacy in Machine Learning
  • BIML Interactive Machine Learning Risk Framework
  • FTC’s “Start with Security” guidelines
  • Mitre Adversarial Threat Matrix
  • NIST Computer Security Resource Center
  • The Security of Machine Learning
  • Proposals for model vulnerability and security

• Machine Learning Attacks and Countermeasures:

  • Membership Inference Attacks Against Machine Learning Models
  • Stealing Machine Learning Models via Prediction APIs
  • Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures
  • Hacking Smart Machines with Smarter Ones: How to Extract Meaningful Data from Machine Learning Classifiers
  • Sponge Examples: Energy-latency Attacks on Neural Networks

• Examples of Real-world Attacks:

  • Finance worker pays out $25 million after video call with deepfake ‘chief financial officer’
  • ISIS ‘still evading detection on Facebook’, report says
  • Researchers bypass airport and payment facial recognition systems using masks
  • Slight Street Sign Modifications Can Completely Fool Machine Learning Algorithms
  • These students figured out their tests were graded by AI — and the easy way to cheat