Machine Learning Security: Threats, Attacks, and Evaluation
Gain insights into analyzing, defending, and evaluating ML systems against adversarial threats. Delve into threat models, metrics, and research-focused evaluation to ensure robust ML security.
Dr. Khayyam Hashmi is a seasoned computer scientist and Generative AI and Machine Learning specialist. As a researcher, educator, and the VP of Technical Content at Educative, he is dedicated to making complex computing concepts accessible for learners worldwide. With a PhD in Computer Science from Wayne State University and academic publications in service-oriented systems and intelligent computing, Khayyam excels at fusing an academic perspective with practical software engineering insights. Khayyam’s courses, writing, and research spans technical topics including adaptive AI systems, neural networks, Intelligent tutoring systems, and production-ready software design.
Join 3.1 million developers at
What You'll Learn
- Draw a system boundary around an ML pipeline and build a structured threat model naming assets, attacker goals, access, and constraints.
- Distinguish ordinary error and distribution shift from deliberate poisoning, backdoors, and evasion attacks.
- Evaluate extraction and privacy risk, including how API design choices change what an attacker can learn.
- Map the attack surface unique to LLM and agentic systems, including prompt injection and supply chain risk.
- Recognize when a defense only appears effective, such as through gradient masking, versus when it actually holds
- Evaluate ML security research and produce an integrated threat model, attack map, and evaluation plan.
Learning Roadmap
1.
Getting Started
Getting Started
2.
ML Systems, Security Objectives, and Attack Surfaces
ML Systems, Security Objectives, and Attack Surfaces
3.
Threat Modeling for ML Security
Threat Modeling for ML Security
4 Lessons
4 Lessons
4.
Training-Time Attacks: Poisoning and Backdoors
Training-Time Attacks: Poisoning and Backdoors
7 Lessons
7 Lessons
5.
Inference-Time Attacks: Adversarial Examples and Evasion
Inference-Time Attacks: Adversarial Examples and Evasion
4 Lessons
4 Lessons
6.
Confidentiality and Privacy Attacks: Extraction and Membership Inference
Confidentiality and Privacy Attacks: Extraction and Membership Inference
4 Lessons
4 Lessons
7.
LLM, Agentic, and Supply Chain Security
LLM, Agentic, and Supply Chain Security
5 Lessons
5 Lessons
8.
Defenses and Mitigations
Defenses and Mitigations
5 Lessons
5 Lessons
9.
Evaluating and Conducting ML Security Research
Evaluating and Conducting ML Security Research
6 Lessons
6 Lessons
Khayyam Hashmi
Computer scientist and Generative AI and Machine Learning specialist. VP of Technical Content @ educative.io.
Trusted by 3.1 million developers working at companies
Anthony Walker
@_webarchitect_
Evan Dunbar
ML Engineer
Software Developer
Carlos Matias La Borde
Souvik Kundu
Front-end Developer
Vinay Krishnaiah
Software Developer
Built for 10x Developers












Free Resources