Feature Selection for Black Hole Attacks

The security issue is essential and more challenging in Mobile Ad-Hoc Network (MANET) due to its characteristics such as, node mobility, self-organizing capability and dynamic topology. MANET is vulnerable to different types of attacks. One of possible attacks is black hole attack. Black hole attack occurs when a malicious node joins the network with the aim of intercepting data packets which are exchanged across the network and dropping them which affects the performance of the network and its connectivity. This paper proposes a new dataset (BDD dataset) for black hole intrusion detection systems which contributes to detect the black hole nodes in MANET. The proposed dataset contains a set of essential features to build an efficient learning model where these features are selected carefully using one of the feature selection techniques which is information gain technique J48 decision tree, Naïve Bayes (NB) and Sequential Minimal Optimization (SMO) classifiers are learned using training data of BDD dataset and the performance of these classifiers is evaluated using a learning machine tool Weka 3.7.11. The obtained performance results indicate that using the proposed dataset features succeeded in build an efficient learning model to train the previous classifiers to detect the black hole attack.