Botnet detection and feature analysis using backpropagation neural network with bio-inspired algorithms

Botnets has been the major type of cybercrime recently, the amount of infected computers gradually increasing each year. Many companies and schools are often troubled with problems, such as DDOS, phishing, spam, and stealing of personal data, because botnet is constantly changing its network structure, attack patterns and data transmission, making it more and more difficult to detect. In this paper, we proposed some new features to detect the botnet traffic, and we found the best solutions by using feature selection algorithm. These two methods are particle swarm optimisation and genetic algorithms, and by using backpropagation network as the classifier, we evaluate our subset feature on botnet detection that shows high detection rate, and we validate that own manufactured feature packet transmission time of regularity can be adopted, and the accuracy will change with the t-value.