Anomaly-based Detection of Denial of Service via Deep Learning Memetic Trained Modular Network

Internet’s popularity for dissemination of data – has birthed the proliferation of attacks that exploit networks for personal gain. Attackers via social-engineering attacks, gain unauthorized access to a compromised device via subterfuge mode and deny users of network resources. Denial of service (DoS) attack is carefully crafted to exploit high levels of network infrastructures. Our study presents a deep learning scheme to effectively classify between genuine and malicious packets. With benchmark XGBoost, Random Forest, and Decision Tree – our resultant model yields an accuracy 0.9984 and F1 0.9945 to outperform the benchmark XGBoost, RF and DT (with F1 of 0.9925, 0.9881 and 0.9805 – and Accuracy of 0.9981, 0.9964 and 0.9815) respectively. Proposed model correctly classified 13,418 cases with a 0.9984 accuracy and has only 283 cases incorrectly classified. Proposed memetic ensemble effectively differentiates malicious from genuine packets using anomaly-based detection.