Due to the prevalence of cloud computing, the dynamic and complex characteristics of cloud computing environments present considerable security protection challenges, and traditional intrusion detection methods are difficult to cope with large amounts of time-series data and new attack patterns. This paper proposes a deep-learning-based framework that integrates long short-term memory (LSTM) networks with an attention mechanism to improve the detection accuracy and robustness of cloud security in terms of dynamically allocating weight to salient time-step features. The framework includes data acquisition, preprocessing, real-time detection and automatic response modules, and the feasibility of implementing the architecture in a low-latency, low-motorized cloud scenario is verified. The proposed framework provides an effective method for building a cloud security protection system. This approach enables adaptive, scalable, and robust online defense against the changing threats in reality.
Rongguo Fu (Thu,) studied this question.