Survey on Evaluating The Sampling For Intrusion Detection System Using Machine Learning Technique

We secure information either in private or government sector as it has become an essential requirement. System vulnerabilities and valuable information magnetize most attackers’ attention. Traditional approaches that are used for intrusion detection, such as firewalls or encryption are not sufficient to prevent system from all attack types. Subsequently the number of attacks through network and other medium has been increased radically in recent years. Thus efficient intrusion detection is required as a security layer against these malicious or suspicious and abnormal activities. Thus, intrusion detection system (IDS) has been introduced as a security technique to detect various attacks. IDS can be identified by two techniques, namely misuse detection and anomaly detection. Misuse detection techniques can detect known attacks by examining attack patterns, much like virus detection by an antivirus application. However they cannot detect unknown attacks and need to update their attack pattern signature whenever there is new attacks .On the other hand, anomaly detection identifies any unusual activity pattern which deviates from the normal usage as intrusion. Although anomaly detection has the capability to detect unknown attacks which cannot be addressed by misuse detection, it suffers from high false alarm rate. In recent years, an interest was given into machine learning techniques to overcome the constraint of traditional intrusion techniques by increasing accuracy and detection rates. New machine learning based IDS with sampling is used in our detection approach. The advantage of IDS (Intrusion Detection system) can greatly reduce the time for system administrators/users to analyse large data and protect the system from illicit attacks. Improve the performance of IDS and the low false alarm rate.