EVMNet: Classification of DDoS Attacks using ELM-Enhanced Deep Learning Model with Feature Fusion
Namrata Govind Ambekar, Sonali Samal, Surmila Thokchom, Yogita Yogita · 2023
The proliferation of high-speed Internet Service Providers and cloud-based applications has increased network vulnerabilities. Distributed denial of service attacks (DDoS) are considered one of the most hazardous threats due to their ability to disrupt services and applications dependent on the Internet instantaneously. As interconnected devices grow, DDoS attacks become more prevalent and destructive. Intrusion Detection Systems (IDS) are required to identify and prevent these threats. As the Internet becomes more intelligent, it becomes more challenging to employ standard security measures. Therefore, it is crucial to develop effective measures for cybercrime. This paper introduces the EVMNet deep learning architecture for classifying DDoS attacks. It is based on cutting-edge Convolutional neural network (CNN) models, such as VGG 16 and MobileNet V2, with Extreme Learning Machine (ELM) feature fusion. The lightweight MobileNet V2 model can be integrated into devices to identify and prevent DDoS attacks. The objective is to differentiate between Benign traffic and DDoS attacks. The model’s effectiveness was assessed using several metrics, such as testing accuracy (TA), precision, sensitivity, false positive rate (FPR), negative prediction value (NVP), and false negative rate (FNR). The results of the proposed study indicate that the suggested model accomplished an impressive testing accuracy of 93.00%, a precision of 92.66%, a sensitivity of 92.44%, a false positive rate of 8.84%, a negative predictive value of 93.30%, and a false negative rate of 8.00% which surpassed the existing state-of-the-art methods.