Deep learning applications for real-time cybersecurity threat analysis in distributed cloud systems
Saswata Dey, Writuraj Sarma, Sundar Tiwari · World Journal of Advanced Research and Reviews · 2023
The newest shift in operations known as distributed cloud systems have greatly advanced the structure of digital environments by providing the ability to scale, be versatile, and cost effective. However, this evolution has significantly raised the cybersecurity danger levels where new kinds of threats like zero-day, DDoS and insider threats are more acute. Known security architectures for managing large-scale systems are frequently ill-suited to rapidly evolving, high-throughput data generated in such contexts. Comprehensive cyber threat detection and analysis in real time through enhanced pattern match in distributed cloud system is made possible by deep learning (DL). The use of security measures that employ the DL models of CNNs, RNNs, as well as the transformer models for detecting security threats are discussed in this article. Some of the features discussed are data preprocessing for imbalanced datasets, model scalability for cloud implementations, as well as incorporating DL with edge computing for better flow. Based on experimental outcomes, according to the evaluation criteria of accuracy and efficiency, DL models can detect anomalies and identify malware earlier, and effectively prevent potential intrusion with higher efficiency than traditional methods. The study also looks at some of the issues with the model; for instance, interpretability; latency; and the need for high-quality data on a big scale. Therefore, it only points to possible further developments using federated learning, privacy-preserving approaches, and multi-model systems to improve threat evaluation in intricate clouds. Thus, this research proves the significance of deep learning in the protection of distributed-cloud systems and brings the gap between idea and application of new approaches to real systems