Deep Radial Intelligence with Cumulative Incarnation approach for detecting Denial of Service attacks

Abstract

Nowadays Denial of Service (DoS) attack is one of the security threats that make the online services unavailable for legitimate users. Therefore, a DoS attack detection system is needed to protect online services against malicious activities. Machine learning approaches are widely used to detect cyber-attacks. The lacunae in the existing machine learning based attack detection systems are more learning time due to vanishing gradient and getting stuck in the local minima due to selection of random weights. In this paper, these issues have been addressed and Deep Radial Intelligence (DeeRaI) with Cumulative Incarnation (CuI) approach is proposed to detect the DoS attacks. The proposed DeeRaI approach learns the intelligence extracted from the radial basis function with different levels of abstraction. The proposed CuI optimizes the weights of the DeeRaI network in which the knowledge gained is progressed to the next generation. Experiments were conducted on benchmark datasets and the proposed approach is compared with existing classifiers and state-of-the-art attack detection systems. It is seen from the performance evaluation that the proposed approach gives promising results than the other existing approaches. Further, it is evident that the proposed approach converges faster and provides best weights compared to the existing optimization methods.