1.1 Intrusion Detection
Research and applications of intrusion detection techniques has resulted in its classification into two-category [1]: misuse intrusion detection and anomaly intrusion detection.
Misuse Intrusion Detection seeks to discover intrusions by precisely defining them ahead of time and watching for their occurrence [2]. For Example, many well known attacks can be discovered by searching for distinguishing patterns or events in the audit trails. The main shortcoming of misuse detection is that future attacks cannot be predicted or detected without hard-coding them into the IDS attack database.
Anomaly Intrusion Detection is based on the assumption that misuse or intrusive behaviour deviates from normal system use. In general, most anomaly detection systems learn a normal system activity profile, and then flag all system events that statistically deviate from this established profile. The strengths of anomaly detection is the ability to abstract information about the normal behaviour of a system and detect attacks regardless of whether or not the system has seen them before and the anomaly detection method can also detect unknown intrusions. Most behaviour models are built using metrics that are derived from system measures such as CPU usage, memory usage, number and time of login, network activity, etc. However, it creates very large overhead for the host machine, which must have the capacity to record all users’ activities to create users’ profiles base on define measure for intrusions [2]. The main weakness of anomaly detection system is their vulnerability to an intruder who breaches the system during their learning phase. A savvy intruder can gradually train the anomaly detector to interpret intrusive events as normal system behaviour. Recent, research projects have addressed a new type of anomaly detection method, which monitor not all user activities but only the privileged processes [5, 6]. This approach is effective
because it eliminates the method also monitors only the privileged processes, which lightens the load of monitoring.