Privacy and security vulnerabilities in edge intelligence: An analysis and countermeasures

Abstract

Recent advancements in deep learning have significantly accelerated the growth of artificial intelligence (AI) technologies, powering applications like the Metaverse, augmented reality (AR), virtual reality (VR), and tactile communications on emerging 6G networks. The proliferation of Internet of Things (IoT) devices and mobile computing has connected vast numbers of devices to the internet, generating enormous amounts of data at the network edge.

To harness the potential of this big data, extending AI capabilities to the network edge has become increasingly vital. Edge AI, or edge intelligence (EI), enables computing tasks to be performed closer to data sources, reducing latency and enhancing efficiency. However, this shift has amplified privacy concerns due to increased data sharing, compounded by the growing prevalence of data breaches. Research also reveals that sharing AI models instead of raw data does not fully safeguard privacy, as certain attacks can still compromise sensitive training information.

This paper reviews Edge Intelligence with a focus on privacy and security issues, identifying critical challenges and vulnerabilities in edge and cloud computing environments. It provides a comprehensive analysis of state-of-the-art solutions to address these concerns, offering valuable insights into enhancing privacy and security in distributed computing systems.