Cybersecurity Threat Detection in SDN Clouds Using Attention Mechanism in RNN

The Cloud Computing (CC) has emerged as the most affordable method of providing big data and AI services to customers online in recent years. Many cloud users often experience serious flaws related to security and privacy, despite the release of various security solutions and patches. The vulnerability of cloud is on the top of Cybersecurity threats and trends list released this year. Economic Denial of Sustainability (EDoS) attack is a type of cybersecurity threat, which is new to cloud environment, gradually increases resource utilisation, which leads to overcharging the cloud customer for the incurred additional cost. We propose an effective method for preventing denial-of-service (DoS) in SDN-based cloud computing environments. In this paper, an enhanced method that uses Attention Mechanism of Recurrenet Neural Network (RNN) for detecting and mitigating EDoS attacks. The main objective is to predict values linked to a cloud user’s resource usage (CPU load, memory usage, etc.). Additionally, in this work, we use deep learning, which is more effective than earlier recommendations, which relied on a predetermined threshold to classify the outliers accountable for the occurrence of frequent errors. The proposed technique offers a self-adjusting threshold, whereas existing systems frequently use a rigid threshold to evaluate the irregularities, leading to increasing mistake rates. The obtained results proved its superiority competing machine learning-based approaches. Our novel method has achieved 96.7% of accuracy in detecting EDoS attacks through extensive testing under a variety of EDoS attacks.