Secure Detection Model Using Black Widow Optimized Features with Bidirectional Learning in Cloud and Fog Network

The Internet of Things, fog, and cloud computing technologies are integrated to provide an effective large-scale computing infrastructure for data-intensive and compute-intensive tasks. Nevertheless, such networks are becoming more susceptible to different intrusions owing to their intrinsically interlinked structure and their extensive use of large-scale network devices. Securing these systems against threats is crucial to ensure trust for end users and protect private information. Recently, Intrusion Detection Systems have been adopted to strengthen security by detecting malicious behavior. Yet, the current attack detection methods suffer from several limitations, such as lower detection accuracy, higher dimensionality, lower computational efficiency, and overfitting issues. Thus, an effective security framework is essential to safeguard against evolving threats in the realm of the Internet of Things, fog, and cloud computing. This research work designed an innovative Deep Learning-based detection methodology for accurate threat detection. The proposed study designed a self-adaptive learning black widow optimization-based rough set theory algorithm for optimal feature selection. This algorithm is deployed to reduce the higher dimensionality of features and computational complexity by selecting significant features. This proposed model adopted a Bidirectional Long Short-Term Memory technique to examine data sequences in both directions, enabling it to capture underlying contextual and temporal relationships within the data. This dual processing enhances the model's ability to identify patterns and anomalies that may indicate an attack. To validate the effectiveness of the proposed framework, comprehensive testing was conducted using UNSW-NB15 and NSL-KDD datasets, along with multiple evaluation criteria. This analysis reveals that the proposed method delivers more accurate and reliable detection outcomes than existing solutions.