State of the Model: Security, Privacy, & Trustworthiness in High-Stakes Machine Learning
Dr. Birhanu Eshete evaluates the trustworthiness of machine learning models in healthcare, autonomous driving, and finance, analyzing adversarial attacks and fairness pitfalls.
Summary
The High-Stakes Proliferation of Machine Learning
Machine learning models are rapidly deployed in critical decision-making systems—including autonomous vehicles, clinical healthcare diagnostics, and algorithmic credit scoring.
Analyzing Security and Privacy Pitfalls in Modern ML
Dr. Birhanu Eshete categorizes core threat vectors—such as adversarial example perturbation, training data poisoning, and model inversion attacks.
Evaluating Transparency, Explainability, and Algorithmic Fairness
The author explores how uncurated training data perpetuates demographic biases and examines technical metrics for verifying algorithmic fairness and model transparency.
Roadmap for Trustworthy and Resilient AI Deployments
The presentation outlines defense-in-depth frameworks, differential privacy mechanisms, and robust verification protocols for deploying dependable AI systems.
Related talks
AI Village·58 min
Journal Club Live
Discusses practical methodologies for discovering exploitable bugs, implementation flaws, and attack surfaces in ML systems.
AI Village·30 min
Erick Galinkin
A primer breaking down core vocabulary, concepts, and foundational theory in machine learning security.
Main Stage·29 min
Ariel Herbert-Voss
Provides a structured adversarial ML red team methodology for attacking production AI systems beyond naive input perturbation.