Enhanced Cyberthreat Detection and Classification with CMCOADL-TDC Using Deep Learning and Optimization Techniques

Cyberthreat detection and classification are crucial fields aiming to develop intelligent systems capable of real-time identification and categorization of various cyberthreats such as malware, phishing, social engineering, and ransomware. Detection involves monitoring networks and systems for suspicious activities like unusual traffic patterns, unauthorized access attempts, and abnormal behaviours. Classification entails identifying specific threat types like viruses, Trojans, or worms, requiring a deep understanding of their characteristics and behaviours. This paper introduces a novel approach, the Cauchy-Mutation Coyote Optimization Algorithm with the Deep Learning Enabled Threat Detection and Classification (CMCOADL-TDC) technique. The proposed technique focuses on accurately identifying cyberthreats through pre-processing, feature selection, and classification. The CMCOADL-TDC technique utilizes a feature selection method based on the Cauchy-Mutation Coyote Optimization Algorithm (CMCOA) model to select optimal feature subsets. A bidirectional gated recurrent unit (BiGRU) model is employed for detection and classification. The BiGRU model's Parameter tuning is performed using the Sunflower Optimization (SFO) model. Additionally, network defense mechanisms are enhanced by employing the time-inhomogeneous hidden Bernoulli model (TI-HBM). To demonstrate its efficacy, extensive simulations were performed by comparing the CMCOADL-TDC approach against state-of-the-art models, showing superior performance. The performance validation of the CMCOADL-TDC approach portrayed a superior accuracy value of 95.58% over existing models.