Suppressing false alarms of intrusion detection using improved text categorization method

Abstract

Although some text processing techniques can be employed to intrusion detection based on the characterization of the frequencies of the system calls executed by the privileged programs, and achieve satisfactory detection accuracy, high false alarms make it hardly practicable in real life. We modify the traditional weighting method tf-idf for suppressing false alarms by considering the necessary information between the processes and sessions. Preliminary experiments with 1998 DARPA BSM auduit data show that our modified method can suppress high false alarms effectively while maintaining satisfactory detection accuracy, which thus make text categorization approaches more practicable for intrusion detection.