Anomaly detection on the edge

Anomaly detection is the process of identifying unusual signals in a set of observations. This is a vital task in a variety of fields including cybersecurity and the battlefield. In many scenarios, observations are gathered from a set of distributed mobile or small form factor devices. Traditionally, the observations are sent to centralized servers where large-scale systems perform analytics on the data gathered from all devices. However, the compute capability of these small form factor devices is ever increasing with annual improvements to hardware. A new model, known as edge computing, takes advantage of this compute capability and performs local analytics on the distributed devices. This paper presents an approach to anomaly detection that uses autoencoders, specialized deep learning neural networks, deployed on each edge device, to perform analytics and identify anomalous observations in a distributed fashion. Simultaneously, the autoencoders learn from the new observations in order to identify new trends. A centralized server aggregates the updated models and distributes them back to the edge devices when a connection is available. This architecture reduces the bandwidth and connectivity requirements between the edge devices and the central server as only the autoencoder model and anomalous observations must be sent to the central servers, rather than all observation data.