Deep Generative Models in the Industrial Internet of Things: A Survey

Abstract

Advances in communication technologies and artificial intelligence are accelerating the paradigm of industrial Internet of Things (IIoT). With IIoT enabling continuous integration of sensors and controllers with the network, intelligent analysis of the generated Big Data is a critical requirement. Although IIoT is considered a subset of IoT, it has its own peculiarities in terms of higher levels of safety, security, and low-latency communication in an environment of critical real-time operations. Under these circumstances, discriminative deep learning (DL) algorithms are unsuitable due to their need for large amounts of labeled and balanced training data, uncertainty of inputs, etc. To overcome these issues, researchers have started using deep generative models (DGMs), which combine the flexibility of DL with the inference power of probabilistic modeling. In this article, we review the state of the art of DGMs and their applicability to IIoT, classifying the reviewed works into the IIoT application areas of anomaly detection, trust-boundary protection, network traffic prediction, and platform monitoring. Following an analysis of existing IIoT DGM implementations, we identify challenges (i.e., weak discriminative capability, insufficient interpretability, lack of generalization ability, generated data vulnerability, privacy concern, and data complexity) that need to be investigated in order to accelerate the adoption of DGMs in IIoT and also propose some potential research directions.

Situational Awareness for Industry 5.0 and Mobile Robotics: A Review

“…Security assessment and development of solutions for mission-critical applications such as vehicle-to-everything connectivity that enables extended situational awareness via use of sensors of other connected vehicles is an active research area [50], [51]. In industrial IoT applications requirements include scalability to large-scale networks, low latency, and advanced and robust security measures [52]. Satellite communications is a way to share SitAw data even when the terrestrial infrastructure does not exist or is destroyed.…”

A comprehensive review of generative adversarial networks: Fundamentals, applications, and challenges

“…Generally, there are different learning types of generative models like Gaussian mixture models (GMM; Reynolds, 2009), hidden Markov models (HMM; Eddy, 1996), latent Dirichlet allocation (LDA; Blei et al, 2003), Boltzmann machines (BM; Ackley et al, 1985), deep belief networks (DBNs; Hinton, 2009), fully visible belief networks (FVBNs; Frey, 1998), VAE (Doersch, 2016), and GANs (Goodfellow et al, 2014). There are various literature reviews providing comprehensive surveys on these different types of the generative models (e.g., Abukmeil et al, 2021;De et al, 2022;Goodfellow, 2017;Harshavardha et al, 2020;Strokach & Kim, 2022;Tomczak, 2022;Turhan & Bilge, 2018).…”

How much BiGAN and CycleGAN-learned hidden features are effective for COVID-19 detection from CT images? A comparative study

“…Most works on DNNs do not pay much attention to the complexity of these models. On the other hand, probabilistic models typically rely on statistical hypothesis tests, which are more simple to implement through the computation of suitable distances in the latent space [42]. The actual capability of GANs to generate data makes them attractive for anomaly detection under two perspectives [43].…”