PERFORMANCE ANALYSIS COMPARISON ON VARIOUS CYBER-ATTACK DATASET BY RELATING A DEEP BELIEF NETWORK MODEL ON AN INTRUSION DETECTION SYSTEM (IDS)

Abstract

In latest eon there are prompt development in Deep learning techniques which is the subset of Machine learning and AI which leads to instinctive innovation of growth in technologies. In general not only in Engineering, researches but also major improvement in medical fields. So deep learning techniques is being applied on emphasizing network security applications.Most observed in network security systems are intruders, which can be viruses,Dos attacks and Penetration among the network makes the difference in the activities of networks.So dynamic methods can be followed to detect and prevent the attack by intruders.In terms, intrusion detection system(IDS) has so many static datasets which was analyzed for traffic alignments. In that aspects for more accuracy and analyzing the deep learning techniques.IDS are classified such as pattern based intrusion detection,time interval based intrusion detection.We focus on antivirus related signature based intrusions. The datasets such as KDDcup 99 and UNSW-NB15 are the pre-existing databases that have variety of patterns.Main focus on generating the False alarm rate ( FAR) and nominal IDS using Deep Belief Network (DBF). This DBF identifies the unpredictable and unanticipated cyber-attacks in both static and dynamic methods. A performance analysis using malware IDS datasets such as KDD dataset DARPA/KDDcup, ADFA-LD and NSL-KDD, CICIDS2017,Iot Device Network logs and UNSW-NB15 features are identified and passed into hidden layers by applying a softmax classifier