Secure and resilient distributed machine learning under adversarial environments

With a large number of sensors and control units in networked systems, the decentralized computing algorithms play a key role in scalable and efficient data processing for detection and estimation. The well-known algorithms are vulnerable to adversaries who can modify and generate data to deceive the system to misclassify or misestimate the information from the distributed data processing. This work aims to develop secure, resilient and distributed machine learning algorithms under adversarial environment. We establish a game-theoretic framework to capture the conflicting interests between the adversary and a set of distributed data processing units. The Nash equilibrium of the game allows predicting the outcome of learning algorithms in adversarial environment, and enhancing the resilience of the machine learning through dynamic distributed learning algorithms. We use Spambase Dataset to illustrate and corroborate our results.