A sliding window technique for covariance matrix to detect anomalies on stream traffic

Along with the development of technology at this era, the need of Internet access service as a media of communication is increasing. This increasing led to anomalies in network traffic. These anomalies, can occur because of a Distributed Denial of Service (DDoS) that deliberately. The impact of an anomaly is to make the user cannot access the Internet service. If left alone, these anomalies can be detrimental to a more parties, both in terms of users and providers of internet access services. Therefore, further research is needed to detect anomalies. The anomalies can be detected by using a covariance matrix. The amount of data that is tested by covariance matrix is often a bottleneck in time, to the use of sliding window to be able to cope with the large number of data. Upon obtainment covariance matrix, then the next step for anomaly detection method is using decision tree to determine the types of anomalies. The test results obtained by using homogeneous test and heterogeneous test is the obtainment of output types of anomalies, and the output can be calculated from the value of the accuracy in detecting (detection rate) and the value of detection errors (false positive rate). A great anomaly detection method able to detect anomalies with parameter values with a high detection rate and low false positive rate.