DeSVig: Decentralized Swift Vigilance Against Adversarial Attacks in Industrial Artificial Intelligence Systems

Abstract

Individually reinforcing the robustness of a single deep learning model only gives limited security guarantees especially when facing adversarial examples. In this paper, we propose DeSVig, a Decentralized Swift Vigilance framework to identify adversarial attacks in an industrial artificial intelligence systems (IAISs), which enables IAISs to correct the mistake in a few seconds. DeSVig is highly decentralized, which improves the effectiveness of recognizing abnormal inputs. We try to overcome the challenges on ultra-low latency caused by dynamics in industries using peculiarly-designated mobile edge computing and generative adversarial networks (GANs). The most important advantage of our work is that it can significantly reduce the failure risks of being deceived by adversarial examples, which is critical for safety-prioritized and delay-sensitive environments.In our experiments, adversarial examples of industrial electronic components are generated by several classical attacking models. Experimental results demonstrate the DeSVig is more robust, efficient, and scalable than some state-of-art defences.

Learning-in-the-Fog (LiFo): Deep Learning Meets Fog Computing for the Minimum-Energy Distributed Early-Exit of Inference in Delay-Critical IoT Realms

“…Hence, like our proposal, DeSVig exploits proximate Fog nodes for providing reliable resource-augmentation with low latencies. However, unlike our contribution, the authors of [20] do not address the topics of the adaptive resource allocation and resource re-configuration.…”

“…Finally, the recent contribution in [20] proposes DeSVig, a decentralized swift vigilance framework for the detection of adversarial attacks in artificial intelligence-empowered industrial applications. Interestingly, by leveraging a twotier decentralized architecture which exploits proximate virtualized Fog servers for resource augmentation, DeSVig allows the reliable detection of abnormal inputs with submillisecond latencies by exploiting the mining capability of suitably designed Generative Adversarial Networks (GANs).…”

“…DeSVig [20] Design and test a Fog-supported decentralized swift vigilance framework for the GAN-enabled detection of cyber-attacks with sub-millisecond delays.…”

“…First, the analysis reported, for example, in [33] supports the conclusion that the scaling K-factors in Eqs. (20) and (24) depend on both: (i) the number of cores n core ≥ 1 equipping each virtual processor; and, (ii) the corresponding fraction r core ∈ [0, 1] of the dynamic power shared by all cores for common operations, as shown below:…”

Machine Learning Threatens 5G Security

“…MEC can be utilized to distribute security functionality close to the end-users and access networks. To detect malicious inputs against deep learning, a distributed approach has been proposed [126] for recognizing adversarial examples. The approach decouples deep learning located in network traffic forwarding elements from the conditional generative adversarial network which is located in the mobile edge.…”