International Journal for Recent Development in Science and Technology

In this paper the author evaluates the performance of two supervised machine learning algorithms such as SVM (Support Vector Machine) and CNN (Convention Neural Networks).Machine learning and Deep learning algorithms will be used to detect whether request data contains normal or attack (anomaly) signatures.Nowadays all services are available on the internet and malicious users can attack client or server machines through this internet and to avoid such attack request IDS (Network Intrusion Detection System) will be used, IDS will monitor request data and then check if it contains normal or attack signatures, if contains attack signatures then request will be dropped.IDS will be trained with all possible attack signatures with machine learning algorithms and then generate train model.Whenever new request signatures arrive then this model applied on new request to determine whether it contains normal or attack signatures.In this paper we are evaluating the performance of two machine learning algorithms such as SVM and CNN and through experiment we conclude that CNN outperform existing SVM in terms of accuracy.To avoid all attacks IDS systems have developed which process each incoming request to detect such attacks and if request is coming from genuine users, then only it will forward to server for processing, if request contains attack signatures, then IDS will drop that request and log such request data into dataset for future detection purpose.To detect such attacks IDS will be prior to train with all possible attack's signatures coming from malicious user's request and then generate a training model.Upon receiving the new request IDS will apply that request on that train model to predict whether its class or whether request belongs to normal class or attack class.To train such models and prediction various data mining classification or prediction algorithms will be used.In this paper the author is evaluating performance of SVM and CNN.In this the author has applied Correlation Based and Chi-Square Based feature selection algorithms to reduce dataset size.This feature selection algorithm removed irrelevant data from dataset and then used model with important features, due to this features selection algorithms dataset size will reduce and accuracy of prediction will increase.