Anomaly Detection Using Support Vector Machines for Time Series Data

Abstract

Research on anomaly detection is of great interest in machine learning and data mining. Detecting anomalies or finding outliers involves identifying abnormal or inconsistent patterns in a dataset. Abnormal data often results from unauthorized activity. Credit card fraud offers a well-known example. Transactions with a stolen or fake credit card can produce suspicious data. A fake card can be made by copying information from an authorized card and using it to create a new unauthorized one. Data such as personal identifying information may be obtained through phishing or from employees who work in credit card companies Another source of abnormal data may derive from unauthorized intrusions in networks. Abnormal traffic or user actions are common signs of intrusions, which may occasion breaches of sensitive or confidential data. Intrusions may also cause sensor networks to generate erroneous data. When a sensor malfunctions, it is unable to capture data correctly and thus may produce anomalies. Abnormal changes in data sources may also result in anomalies