ANN AND VOTING BASED NOVEL APPROACH FOR BUILDING BALANCED IDS

Today, the increasing number of network attacks requires the development of high-precision attack detection mechanisms in the field of cyber security. Many organizations have intrusion detection systems (IDS) that play an important role in detecting and preventing attacks on their networks, whether they are signature, anomaly detection, or hybrid. This research proposes a new way to improve the performance of IDS by solving the main problems such as false alarms, feature selection and imbalance of neural network architecture. The proposed method uses a majority function-based and improved voting mechanism to select the optimal features among several algorithms such as Anova FTest, Recursive Feature Elimination, and Cross Validation Recursive Feature Elimination. At the same time, it is suggested to use the entropy-based feature selection method combined with the information gain ratio method to improve the feature selection process. The proposed method includes a real-time artificial neural network topology balancing algorithm, as well as an optimization algorithm for the number of hidden layers and hidden neurons.