Cyberattack Prevention in Cloud-Based Systems Using Butterfly Optimization Algorithm BOA Based CNN for Intrusion Detection

Abstract

Cloud computing offers significant benefits in scalability and flexibility but remains highly vulnerable to cyber threats such as malware injections, DDoS attacks, and data breaches. Such ever-growing attacks are even more difficult for traditional IDS and thus end up with higher false positives, which is much typical of their low adaptability. Deep learning-based IDS, an especially CNN-is expected to do really well in fighting the very difficult signatures of attacks. The disadvantages of CNN involve very long periods, as well as the exhaustive computational resource for extensive hyperparameter tuning. This paper hence contributes by proposing a butterfly optimization-based hyperparameter fine-tuning of CNN-based intrusion detection systems. Further, hybridization of BOA and CNN can enhance detection accuracy; reduce false alarms; and better adaptability to real-time emerging threats in cloud environments. The proposed framework has been evaluated using the KDD-99 dataset and achieved an overall 98.9% reliability, including 98.35% F1 score, 98.55% recall, and 98.75% consistency. Thus, the achieved results speak for how effectively the BOA optimization can improve the performance of the CNN in the identification of disturbances. So, the study provides intelligent and scalable solutions for IDS and as associated with increases detection of threats in real time but covering low computational overheads. Future work would thus apply this methodology to a broader set of datasets, as well as incorporate other metaheuristic algorithms for better detection optimization into the scope of the work.