Low-rank block sparse decomposition algorithm for anomaly detection in networks

In this paper, a method is suggested for the anomaly detection in wireless networks. The main problem that is addressed is to detect the malfunctioning sub-graphs in the network which bring about anomalies with block sparse structure. The proposed algorithm is detecting the anomalies considering the low-rank property of the data matrix and the block-sparsity of the outlier. Hence, the problem boils down to a compressed block sparse plus low rank decomposition that is solved with the aid of the ADMM technique. The simulation results indicate that the suggested method surpasses the other technique especially for higher block-sparsity rates.