Attacks against Machine Learning Models in 5G Networks

Abstract

Artificial intelligence and machine learning are revolutionising almost every industry with a seemingly endless list of applications. This list also includes mobile networking for which employing machine learning algorithms can improve efficiency, latency, and reliability of services and applications. For this reason, functionality of future 5G networks is expected to depend on accurate and timely performance of its artificial intelligence components, and disturbance in the functionality of these components may have negative impact on the entire network. This study focuses on analysing adversarial example generation attacks against machine learning based frameworks that may be present in the next generation mobile networks. In particular, we study transferability of adversarial example attacks to the 5G domain and evaluate their negative impact on the network performance in several realistic use case scenarios.