Anomaly Detection Using System Call Sequence Sets.

This paper discusses our research in developing a generalized and systematic method for anomaly detection. The key ideas are to represent normal program behaviour using system call frequencies and to incorporate probabilistic techniques for classification to detect anomalies and intrusions. Using experiments on the sendmail system call data, we demonstrate that concise and accurate classifiers can be constructed to detect anomalies. An overview of the approach that we have implemented is provided.Cochin University of Science and TechnologyJOURNAL OF SOFTWARE, VOL. 2, NO. 6, DECEMBER 200