Bounding an Attack's Complexity for a Simple Learning Model

Provided by: University of California San Francisco
Topic: Security
Format: PDF
As machine learning becomes more prevalent as a systems and networking analysis and detection tool, it is becoming an attractive target for attackers who seek to manipulate the system. The authors examine a naive model for assessing the effectiveness of classifiers against threats poised by adversaries determined to subvert the learner by inserting data designed for this purpose. Based on this model, they analyze the attack in detail, develop bounds on the adversary's capability, and discuss the implications for the security of learning-based detection systems.

Find By Topic