Strengthening intrusion detection techniques through emerging patterns

Abstract

In todays world where nearly every company is dependent on the Internet to survive, it is not surprising that the role of intrusion detection has become extremely important within the last decade. Intrusion detection involves determining whether some entity has attempted to gain, or worse, it has gained unauthorized access to the system. The task of current intrusion detection systems is detect possible threats not only from insiders but also from outsiders. Based on our current knowledge, there are two things the system administrator could do in order to keep secure his system. First, use preventive measures. Second, make use of the audit logs. Due to the sheer volume of the logs, it is required that this task be performed automat- ically. Data Mining eld of study has help to partially automatize this process. However, the current state of art has still left too much to the administrator and sometimes it distract the administrator raising false alarms. This work propose to apply a new technique, successfully used in others elds of knowledge as Bioinformatics and Classi cation Systems, in order to de ne more accurately user's pro les and to detect more intruders, raising a lower number of false alarms and having a precision higher than other techniques.