Fractal Based Anomaly Detection over Data Streams

• Published in: Web Technologies and Applications pp. 550 - 562
• Main Authors: Gong, Xueqing, Qian, Weining, Qin, Shouke, Zhou, Aoying
• Format: Book Chapter
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg
• Series: Lecture Notes in Computer Science
• Subjects: Anomaly Detection; Data Streams; Fractal
• ISBN: 364237400X; 9783642374005
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-642-37401-2_54

Robust and efficient approaches are needed in real-time monitoring of data streams. In this paper, we focus on anomaly detection on data streams. Existing methods on anomaly detection suffer three problems. 1) A large volume of false positive results are generated. 2) The training data are needed, and the time window of appropriate size along with corresponding threshold has to be determined empirically. 3) Both time and space overhead is usually very high. We propose a novel self-similarity-based anomaly detection algorithm based on piecewise fractal model. This algorithm consumes only limited amount of memory and does not require training process. Theoretical analysis of the algorithm are presented. The experimental results on the real data sets indicate that, compared with existing anomaly detection methods, our algorithm can achieve higher precision with reduced space and time complexity.