07. References & Standards

This directory lists authoritative benchmarks, regulatory frameworks, foundational academic literature, and open-source toolkits for Machine Learning Security and Adversarial AI engineering.

---

1. Security Standards & Risk Taxonomies

• NIST AI 100-2: Trustworthy and Responsible AI — Official NIST taxonomy of attacks and mitigations in Artificial Intelligence and Machine Learning.
• MITRE ATLAS™ (Adversarial Threat Landscape for Artificial Intelligence Systems) — Globally accessible knowledge base of adversary tactics, techniques, and case studies against AI-enabled systems.
• OWASP Machine Learning Security Top 10 — Community-driven standard identifying the top security vulnerabilities in machine learning deployments.
• ISO/IEC 42001:2023 AI Management System — International management system standard providing requirements for establishing, implementing, and continually improving AI security governance.

---

2. Landmark Academic Research Papers

Adversarial Evasion & Robustness

• Goodfellow, I. J., Shlens, J., & Szegedy, C. (2014). Explaining and Harnessing Adversarial Examples. arXiv preprint arXiv:1412.6572.
  (Introduced the Fast Gradient Sign Method (FGSM) and linear model hypothesis).
• Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu, A. (2017). Towards Deep Learning Models Resistant to Adversarial Attacks. arXiv preprint arXiv:1706.06083.
  (Introduced PGD iterative attack and minimax adversarial training formulation).
• Carlini, N., & Wagner, D. (2017). Towards Evaluating the Robustness of Neural Networks. IEEE Symposium on Security and Privacy (SP).
  (Formulated C&W L2/L_inf optimization attacks bypassing defensive distillation).

Data Poisoning & Trojan Backdoors

• Gu, T., Dolan-Gavitt, B., & Garg, S. (2017). BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain. IEEE Access.
  (First paper demonstrating backdoor Trojan trigger injection in neural network weights).
• Shafahi, A., Huang, W. R., Najibi, M., Suciu, O., Studer, C., Dumitras, T., & Goldstein, T. (2018). Poison Frogs! Clean-Label Backdoor Attacks on Machine Learning. NeurIPS.
  (Demonstrated targeted clean-label feature collision data poisoning without altering training labels).
• Tran, B., Li, J., & Madry, A. (2018). Spectral Signatures in Backdoor Attacks. NeurIPS.
  (Introduced SVD feature representation filtering to detect and remove backdoor training samples).

Model Intellectual Property Theft & Privacy Leakage

• Tramèr, F., Zhang, F., Juels, A., Reiter, M. K., & Ristenpart, T. (2016). Stealing Machine Learning Models via Prediction APIs. USENIX Security Symposium.
  (Pioneered black-box model extraction attacks against commercial prediction APIs).
• Shokri, R., Stronati, M., Song, C., & Shmatikov, V. (2017). Membership Inference Attacks Against Machine Learning Models. IEEE Symposium on Security and Privacy (SP).
  (Formulated shadow model techniques to infer individual record membership in training sets).

---

3. Open-Source Security Toolkits & Frameworks

Tool / Repository	Organization	Link	Purpose
Adversarial Robustness Toolbox (ART)	Linux Foundation	GitHub	Comprehensive python library for ML security evaluation and defense.
Microsoft Counterfit	Microsoft Azure	GitHub	Automation tool for black-box penetration testing of ML models.
Safetensors	Hugging Face	GitHub	Safe, zero-copy tensor serialization format replacing unsafe pickle files.
Picklescan	Hugging Face	GitHub	Security scanner for detecting dangerous opcodes in Python pickle artifacts.
Modelscan	Protect AI	GitHub	Open-source tool for scanning ML model artifacts for code execution risks.
Fickling	Trail of Bits	GitHub	Decompiler and static analyzer for inspecting Python pickle bytecode.
PyTorch Opacus	Meta AI	GitHub	High-speed library for training PyTorch models with Differential Privacy (DP-SGD).
RobustBench	Community	RobustBench Leaderboard	Standardized benchmark leaderboard for empirical adversarial robustness evaluation.

---

Return to Module Overview & Index →