International Journal of Engineering and Technology

Security is an important challenge in internet based communication. In such a scenario, intrusion detection systems help to secure the data through the identification of normal and abnormal behaviors. In order to model these behaviors accurately and to improve the performance of the intrusion detection system, a temporal oriented heuristic genetic neural network (THGNN) is proposed in this paper. In this model, feature selection, structure design and weight adaptation are jointly in considered to analyze the interdependence of input features which helps to modify the network structure and connection weights. Moreover, the genetic algorithms are proposed to work with input nodes and hidden nodes. The crossover operator based on temporal constraints are introduced and considering the relationship between genotype and phenotype. Moreover, a temporal logic based adaptive mutation rate is applied, and the mutation operation is performed heuristically from time based weight adaptation, node manipulation. When the population is not evolved continuously for a time interval, the mutation rate is increased and the mutation type is changed. This temporal heuristic approach helps to perform weight adjustment effectively. Experimental results obtained using the KDD-99 dataset show that the proposed THGNN achieves better detection accuracy in terms of increased detection rate and decreased false positive rate.