Optimized Network Security Attack Detection Algorithm Based on Deep Deterministic Policy Gradient (DDPG)

As network environments evolve and attack vectors grow increasingly complex, traditional intrusion detection systems (IDS) often struggle to maintain accuracy, adaptability, and realtime responsiveness. This paper proposes a novel intrusion detection framework based on the Deep Deterministic Policy Gradient (DDPG) algorithm, a deep reinforcement learning method designed for continuous action spaces. By modeling intrusion detection as a Markov Decision Process (MDP), the system continuously learns optimal detection strategies through interaction with dynamic network traffic. To enhance performance, the proposed model integrates a multi-objective reward function balancing detection precision and latency, a prioritized experience replay mechanism to address class imbalance, and a deep actor-critic architecture for expressive policy representation. Experimental evaluations conducted on the NSL-KDD and UNSW-NB15 datasets demonstrate that the proposed method significantly outperforms baseline modelsincluding SVM, Autoencoder, DQN, and standard DDPG-in terms of accuracy, F1-score, and AUC.