Adaptive Cyber Threat Detection Using Hybrid Deep Learning Models in Multi-Cloud Environments

Abstract

This increased the complexity and scale of cybersecurity concerns, as multi-cloud is now mainstream, exposing distributed infrastructures to sophisticated and evolving cyber threats. Static signatures and restricted adaptability limit conventional intrusion detection systems (IDS), making it difficult to detect advanced persistent threats and zero-day attacks. Proposed adaptive cyber threat detection framework for multi-cloud in this paper utilizes hybrid deep learning models to detect and response effectively. By employing CNN for extracting spatial features combined with LSTM networks to analyze temporal patterns that will allow discovering known and unknown attack scenarios. Besides, it includes an adaptive learning module to further learn and evolve over time with changing threat intelligence and variation of cloud workloads. Evaluate the framework against benchmark cybersecurity datasets and simulated multi-cloud traffic environment, showcasing its superior detection rates, lower false positive rates, and faster response times compared to traditional or standalone machine learning approaches. The experimental results indicate that the proposed hybrid model offers more than 97% detection accuracy with scalability and robustness over heterogeneous cloud platform. Thus the proposed solution can be used as a smart and scalable defense tool which could secure modern multi-cloud infrastructures against the increasing threat of cyber-attacks.