Adversaries of AI applications continuously advance their tools for evasion, poisoning, extraction, and inference against the underlying machine learning models to achieve their malicious goals. This technical talk will introduce the actual open source tools to reproduce these attacks and the tools needed to defend and evaluate applications before deploying and exposing them to adversaries.

RSA® Conference, the world’s leading information security conferences and expositions, concluded its 30th annual event, which took place as a fully virtual experience this week. This year’s theme was Resilience, and in honor of all that the cybersecurity industry has accomplished over the past three decades, offered a must-attend celebration of thought leadership and education.


More than 20,000 people registered to view 449 sessions including, keynotes, track sessions, interactive programs, tutorials, seminars, and many special digital-first programs that focused on best practices and innovation. The content covered many of the industry’s most pressing topics, including privacy, machine learning and artificial intelligence, analytics, anti-fraud, policy and government, and risk management and governance.

Evasion, Poisoning, Extraction, and Inference: Tools to Defend and Evaluate

Conference: RSA Conference 2021

Abstract

Post a comment

Related work

Adversarial Machine Learning - Learn Why and How to Break AI!

Designing for Autonomous Vehicles at Scale With Hussein Mehanna of Cruise

Malproxying: Leave Your Malware at Home

Smart Green Computing Cloud Native Operations

The Devil is in the GAN: Defending Deep Generative Models Against Adversarial Attacks

Defending the 2018 Midterm Elections from Foreign Adversaries