Anomaly detection in time series via robust PCA

• Published in: 2017 2nd IEEE International Conference on Intelligent Transportation Engineering (ICITE) pp. 352 - 355
• Main Authors: Yongjun Jin, Chenlu Qiu, Lei Sun, Xuan Peng, Jianning Zhou
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-09-2017
• Subjects: Anomaly detection; Matrix decomposition; Principal component analysis; robust PCA; Robustness; Security; sparse and low rank matrix recovery; Sparse matrices; Time series analysis
• ISBN: 9781509062720; 1509062726
• DOI: 10.1109/ICITE.2017.8056937

In the era of big data, massive data has been accumulated in integrated application platform of traffic management of public security. Confidential information leakage is one issue of great importance in the process of data collecting, storing, processing and utilizing. In this work, we propose an anomaly detection approach for the prevention of confidential information leakage. Assume that users' query behavior obeys certain trend and the occurrence of abnormal or malicious behavior is sparse and random distributed. Arrange query series as matrix, the anomaly detection problem can be formulated as a spares and low rank recovery problem. List of suspicious staff is constituted by including user accounts whose sparse recovery has significant nonzero elements. Experimental results on simulated and real data are given, showing the effectiveness of the proposed method.