DeepGuard: fortifying intrusion detection with LSTM and walrus optimization for smart-cities

The development of effective Intrusion Detection Systems (IDS) is essential to protect computer networks given the complexity and sophistication of cyber threats, which are constantly becoming more complicated and sophisticated in smart cities. In this paper, we propose an IDS architecture that combines the power of Long Short-Term Memory (LSTM) models and the optimization capabilities of the Walrus Optimization Algorithm (WOA) to enhance cybersecurity in the smart city.Recurrent neural networks of the LSTM variety are excellent at identifying temporal dependencies in sequential data, which makes them suitable for tasks requiring intrusion detection. WOA offers a metaheuristic optimisation strategy to train the LSTM model and improve its functionality. The UNSW_NB15 dataset, a well-used benchmark dataset for intrusion detection research, is employed in the proposed IDS system. The system's efficiency in identifying and categorising network intrusions is evaluated using accuracy and loss as the performance indicators.It achieves 96% accuracy. The system achieves improved accuracy and robustness by utilising LSTM's capacity to record temporal relationships and utilising WOA's optimisation capabilities. In conclusion, the proposed IDS architecture using LSTM and the Walrus Optimisation Algorithm shows promising potential in accurately identifying network intrusions in smart city applications.