Improving Internet of Things communications through compression and classification

The amount of data produced and exchanged in the Internet of Things is continuously increasing. The associated management costs for information transmission and classification are becoming an almost unbearable burden due to the unprecedented number of data sources and the intrinsic vastness of the dataset. In this paper, we propose a novel lightweight approach capable of alleviating both aspects by leveraging on the advantages offered by classification methods to optimize communications and by enhancing information transmission to simplify data classification. In particular, we propose to adopt Motifs, recurrent features used for signal categorization, in order to compress data streams: in such a way it is possible to achieve compression levels of up to an order of magnitude, while maintaining the signal distortion rate within acceptable bounds and allowing for simple lightweight distributed classification and anomaly detection techniques. We elaborate about data representation and motif extraction methods for constrained devices, proposing a simple and effective solution for the problem. We validate our approach with an extensive simulation campaign thoroughly spanning the system parameter set. This work paves the road ahead for the realization of a universal signal processor for constrained devices in the Internet of Things, which will be capable of appropriately handling any given data while at the same time increasing communication efficiency.