Training MLP neural network to reduce false alerts in IDS

Due to the tremendous growth of the Internet and Network based services, the severity of network based computer attacks have significantly increased. Thus, IDS play a vital role in network security. Intrusion detection system tries to detect computer attacks by examining various data records, log audits etc. Many existing IDS such as Snort are signature based system. The problem with such a system is that it cannot detect novel attacks whose signature is not available and hence generates a high rate of alerts. In this paper Multilayer Perceptron (MLP) with Back-Propagation algorithm is used to classify attacks. We train and test MLP with KDD99 training dataset. We use KDD99 dataset which is a subset of the DARPA dataset. It is a preprocessed dataset and is most suitable for our system. We analyze the working of MLP by performing various experiments. We observed that MLP Neural network requires large training time. Once it trained, detects known as well as unknown attacks and also reduces false alerts.