Hybrid CNN-LSTM Architecture for Robust Cloud Security Through Anomaly Detection and Threat Mitigation

Abstract

Security is a must in the case of the cloud which lately on the contrary we are using cloud computing for multiple applications, there's the need for strong intrusion detection systems to address the problems that are of great concern. The new idea in our publication is a CNN-LSTM hybrid architecture which is designed to minimise the rate of security breaches in cloud computing. Accordingly, the conceptual IASP combines the assets of CNNs that are equipped to recognize the correlation amongst spatial domains in data and LSTMs which are responsible for dependency relations and the formulation of short-term memory. The design of this model is specifically aimed at overcoming the challenges such as periodicity in cloud data, which ensures that it can correctly and timely detect the anomalies. There are studies in which we have been dealing with two strategies to capture seasonality, namely, independent modelling for each season and inclusion of seasonality as a feature into the input data. Our hybrid CNN-LSTM model proposed here shows the superiority in detecting anomalies as collusion with various cloud security scenarios. Our solution improves the accuracy of detection by 15% and the false positive rate by 20% when compared to other existing methods. The model also accomplishes a detection speed that is 30% faster than the slower traditional methods and thus becomes a useful instrument for the protection of cloud systems.