HYBRID NETWORK BASED INTRUSION DETECTION SYSTEMS FOR DoS ATTACKS

Abstract

Intrusion Detection System (IDS) is an essential component in the overall network and data newlinesecurity. With the rapid advancement in network technologies, detection of attacks based on newlinethe analysis of contextual information may be specific to individual applications and newlinenetworks. Such type of problem can be overcome with the help of hybrid IDS. This thesis newlinediscuss about new Hybrid Intrusion Detection System (IDS) models that are provided with newlineadaptive mechanisms to handle and detect large volume of malicious data generated by newlineDenial of Service (DoS) attacks. These attacks are generally based on flooding of packets newlinewith the intention of over filling the victim resources. Today, these attacks are capable of newlineinterrupting the networks of almost any type of size. To address this problem, three models newlinenamely GA-PSO, GA-SVM and PSO-SVM are proposed that provides increased accuracy to newlinedetect intrusion while using hybrid model rather than primary algorithms. These models have newlinethe knowledge of both known and unknown potential malicious activities in the network newlinetraffic and raise an alarm whenever a suspicious activity is detected. Hybrid model uses the newlineapproach of integrating different learning or decision models. Each model works in a newlinedifferent aspect and exploits different set of features. Integrating these models gives better newlineperformance than an individual model. newlineOne of the important research challenges for constructing high performance hybrid IDS is to newlinedeal with large volume of records containing large number of features. Large number of newlinefeatures can make it harder to detect malicious patterns, causing slow training and testing newlineprocess, higher resource consumption as well as poor detection rate by considering this issue newlinein the design of proposed IDS by reducing and removing irrelevant features in the bench newlinemark dataset. Preprocessing performs elimination of out of range values, impossible data newlinecombinations and missing values present in the dataset.