<p>Cloud computing provides convenient, on-demand access to shared computing resources such as networks, servers, storage, applications, and services. These resources can be added or removed quickly without requiring much effort from the service providers. Privacy and security are the major concerns for the success of cloud computing due to its open and distributed architecture, which can be vulnerable to unauthorized access. Certain conventional intrusion detection systems are implemented in internet or intranet settings, which lack scalability and the security policies that are static or rarely change over time. To overcome these limitations, a Deep Maxout Network fused with Deep Long Short-Term Memory (DeepMaxout LSTM) trained by the Taylor Flamingo Search Algorithm (TFSA) is proposed for an Intrusion Detection System (IDS). TFSA is developed in this research to boost the intrusion detection process by tuning the hyperparameters and to determine the optimal parameter of DeepMaxout LSTM. Initially, the input data obtained after cloud simulation is allowed to the Spark framework with slave nodes and master nodes. In the slave node, various processes are carried out. Here, pre-processing is done by Z-score normalization, which is followed by feature fusion carried out by a Deep Belief Network (DBN). These fused features are then fed towards data augmentation by the Synthetic Minority Oversampling Technique (SMOTE). Then, the augmented data is sent to the master node, where the intrusion is detected by DeepMaxout LSTM, which is the combination of Deep Long Short-Term Memory (DLSTM) and Deep Maxout Network (DMN). Also, TFSA is formed by combining the Taylor series and the Flamingo Search Algorithm (FSA). Finally, the analysis of the proposed TFSA_DeepMaxout LSTM is done by various metrics like accuracy, recall, precision, F1 score, computational time, and training time, exhibiting high percentages of 0.909, 0.947, 0.965, 0.956, 0.288&#xa0;s, and 0.198&#xa0;s, respectively.</p>