Abstract - The advent of digital ecosystems today through the use of technologies such as Cloud Computing, Internet of Things (IoT), and Edge Computing has made it possible for us to achieve greater connectivity and scalability, but at the same time, has brought about the complexity of cybersecurity threats in dynamic and diverse network environments due to their vulnerability to advanced attack mechanisms such as Distributed Denial of Service (DDoS) attacks, ransomware, and advanced persistent threats. Conventional intrusion detection systems, including the widely known signature-based and rule-based IDSs, have proven ineffective in detecting such emerging threats since they depend on fixed patterns and rules while anomaly-based IDSs suffer from high rates of false positives. However, recent studies employing Artificial Intelligence techniques, especially Deep Learning techniques such as Convolutional Neural Network and Gated Recurrent Units (GRU) have had promising results for identifying the complex patterns of network traffic. Nevertheless, since they are considered to be black boxes, their use becomes complicated due to lack of model interpretability which is key in cybersecurity since decisions made cannot be trusted nor be compliant with regulations. To solve this problem, Explainable Artificial Intelligence methods such as SHAP and LIME have been developed to help in explaining decisions but are often used individually although there is no harm in using both. This research, therefore, suggests developing a hybrid approach of SHAP and LIME to explain predictions from artificial intelligence models while proposing a novel feature selection approach using the two methods known as SHAP- LIME Aggregated Feature Selection (SLA-FS). The features selected by SLA-FS will be used in Deep Learning models such as Convolutional Neural Network (CNN), GRU, and Hybrid-CNN-GRU model. Experimental results indicate that this proposed approach is better than existing models in terms of accuracy with 99.64% accuracy.