The design of hybrid intrusion detection system by using sine cosine algorithm with support vector machine

Nowadays the use of computer systems became terrifically significant due to advance technology, simplicity and availability of the internet. Moreover, the computer systems play a crucial role for our daily life in various objective such as; storing critical data, online education and online shopping. Nevertheless, users of the computer systems frequently encounter threats which undermines the maintenance and stability of their operating systems. Particularly, politician systems, banks and etc. the intruders attempt to obtain the individual information by unauthorized access to the private systems. To accomplish this vulnerability, majority of scientists have attracted their attention the use of hybridized methods as intrusion detection systems. Therefore, diverse techniques like meta-heuristic algorithms and machine learning methods are proposed in order to boost the performance of the presented methods as well to select minimum attributes. In this thesis, it is goaled to combine two successful algorithms selected from meta-heuristic and machine learning algorithms (Binary Sine Cosine Algorithm (BSCA) and Support Vector Machine (SVM)) to attain superior performance and simultaneously enhance accuracy. It foresaw to use BSCA for feature selection for its success in feature selection and to use SVM algorithm as a classifier due to its success in classification. The main objective this thesis is to highlight the implicit of a new algorithm, which is integrated as a hybrid form machine learning algorithm, in order to detect attackers and to acquire the most acknowledge hybrid intrusion detection systems. Two different data sets were used to evaluate the performance of the presented model.The result of the new proposed hybrid system will be compared to some success rates of other machine learning techniques such as; RBF and Polynomial Support Vector Machine (RBF-SVM and Polynomial SVM), Random Forest (RF), K-nearest neighbor (k-NN), Naive Bayes classifier (NBC), SVM with Binary Particle Swarm Optimization (BPSO-SVM). And some current existing studies (such as IWD-SVM, GA-SVM, MBGW-SVM and LOA-CNN) which is selected from the literature.Using NSL-KDD and UNSW-NB15 datasets, the proposed method achieved 99.30% and 99.70% accuracy respectively.