Intrusion Detection for Cyber Security Systems Using Machine Learning and Data Mining

Abstract

Emerging rate of security threats in the network requires tremendously reliable solution. Many researchers have studied and worked on numerous ways to identify invasions. In my research I have considered the usage of Machine Learning Approach for the phases of Intrusion Detection. Meanwhile, Intrusion Detection System (IDS) have a significant role in the framework and evolution of resolute linkage framework which provide safe and secure system by differentiating and obstructing variety of hazards. Many techniques that are built on Machine Learning Approaches have been established before. Although, those techniques are not effective and efficient enough to detect all types of infringements. In this thesis, I have done comprehensive analysis of numerous Machine Learning Techniques to discover the base of glitches which is related to numerous Machine Learning Techniques in identifying activities of invasion. Limitations related to each Machine Learning Techniques are deliberated in this thesis. This research also contains numerous data mining tools designed for Machine Learning Techniques. Constant standard datasets occur severely to estimate and assist enactment of the Detection Structure. The subsets of numerous datasets are, for instance, ADFA13, DARPA98, ISC2012, KDD99, etc. They are used to evaluate the enactment of Intrusion Detection strategies, but I have used the modern strategies in my thesis, that is, CICIDS2017 which provides more improved accuracy. In this thesis I originated a wide evaluation of the existing datasets by resources of our own estimated criteria and propose an estimate framework for Intrusion Detection System (IDS) datasets. I am ascertain of the challenges of network Intrusion Detection and offered the principles for the future study on glitch detection.