Machine Learning at the Edge
Sundara Mohan, Alok Manke, Shanti Verma, K. Baskar · Advances in information security, privacy, and ethics book series · 2024
In this study, a novel system named “EdgeAnomaly,” is proposed, which leverages generative adversarial networks (GANs) for anomaly detection on wireless sensory networks (WSNs) which are operating at the edge. The proliferation of internet of things (IoT) for devices has led to an exponential increase in data generation by WSNs, necessitating efficient and effective anomaly detection mechanisms. In traditional anomaly detection methods often struggles to cope with the dynamically and diverse nature of WSN data, particularly in resource-constrained edge computing environment. To address these challenges, the employment of GANs, a type of deep learning model capable of generating synthetic data samples resembling the original data distribution. By training the GAN on normal WSN data, EdgeAnomaly will learn to generate representative samples of normal behavior, which enables it to identify deviations indicative of anomalies.