Affiliations: [a] Department of Computer Engineering, D.Y. Patil University Ambi, Talegaon Dabhade, Pune, Maharashtra, India-410507 | [b] Department of Information Technology, Dr. D.Y. Patil’s Ramrao Adik Institute of Technology, Sector 7, Phase I, Pad. Dr D.Y. Patil Vidyapeeth, Nerul, Navi Mumbai, Maharashtra, India-400706 | [c] Department of Computer Engineering, School of Engineering & Technology, D Y PATIL University Ambi, Talegaon Dabhade, Pune, Maharashtra, India-410507
Abstract: Malicious traffic segregation and attack detection caused major financial loss and became one of the most serious security hazards. Moreover, cyber security attack is the major issue, which impacts network security. The network attack methods are constantly being upgraded by the technology development and it remains a major issue for detection and protection against network attacks. For this, it is required to present an effective strategy for detecting and maintaining network security. The work provides timely and accurate congestion attack detection and identification. In the Internet of Things (IoT) cloud system malicious traffic segregation and attack detection based on a hybrid optimization-enabled deep learning (DL) network is developed in this research. At first, the input log files are gathered from the simulation of IoT sensors and the superior route is selected by the proposed Fractional Hunger Jellyfish Search Optimization (FHGJO) algorithm. The FHGJO is the integration of Hunger Game Jelly Fish Optimization (HGJO) and Fractional Calculus (FC). Furthermore, the HGJO is the combination of Hunger Game Search Optimization (HGS) with Jellyfish Optimization (JSO). Then, the segregation is done based on the fitness measures and for preprocessing; the input data is fed using quantile normalization. The feature selection process is employed using the weighted Euclidian distance (WED). With the SpinalNet, the malicious segregation is categorized as malicious and non-malicious and the proposed FHJGO is used to tune the SpinalNet. Furthermore, the proposed FHGJO-trained Deep Quantum Neural Network (DQNN) is utilized to detect the attack and classifies it into a Denial-of-Service (DOS) attack, Distributed Denial of Service (DDoS) attack, and buffer overflow attack. Moreover, the proposed model is evaluated using the NSL-KDD dataset and BoT-IoT dataset. The proposed method ensures network security with 0.931 accuracy, 0.923 sensitivity, and 0.936 specificity.
Keywords: Deep Quantum Neural Network, weighted Euclidian distance, Hunger Game Search Optimization, Jelly Fish Optimization, Fractional Calculus
DOI: 10.3233/WEB-230214
Journal: Web Intelligence, vol. Pre-press, no. Pre-press, pp. 1-23, 2024
