AI-Powered Sparse Temporal Attention Framework for Robust LDAP Attack Detection in IoT Systems

IoT systems are growing fast. With that growth comes more risk, especially from application-layer attacks like Lightweight Directory Access Protocol (LDAP). Many detection models still struggle to detect this attack. They often fail when data is unbalanced and take too long to run. In this context, this work proposed an AI-powered approach. It combines Temporal Sparse Attention with LSTM to detect LDAP attacks. The model uses Forensic-Based Investigation Optimization to selected the optimal features. It also learns meaningful patterns from categorical inputs using embedding drift. The model was tested on the CIC-DDoS2019 dataset. It reached 99.86% accuracy. In addition of that, proposed model outperforms the baseline and state-of-the-art models. Results show strong performance with low complexity. It fits well in real-time IoT systems.