Optimization Strategy and Application of Intrusion Detection System Based on Deep Learning

本研究提出一种基于CNN-LSTM的混合深度学习模型，结合CNN的静态数据特征提取能力和LSTM的时序建模优势，以提升入侵检测性能。讨论基于深度学习的入侵检测系统的优化策略并利用仿真实验对比分析了CNN、LSTM、GRU和CNN-LSTM这几种模型在NSL-KDD数据集上的性能。经过实验验证，CNN-LSTM模型在Accuracy (94.1%)、Recall (92.8%)和F1-score (93.4%)等关键指标上都优于单一模型，且降低了误报率和漏报率，验证了其在复杂攻击模式检测中的有效性，为入侵检测系统的优化提供了可行方案。This study proposes a hybrid deep learning model based on CNN-LSTM, which combines the static data feature extraction ability of CNN and the temporal modeling advantage of LSTM to improve intrusion detection performance. It discusses optimization strategies for intrusion detection systems based on deep learning and compares the performance of CNN, LSTM, GRU, and CNN-LSTM models on the NSL-KDD dataset using simulation experiments. Through experimental verification, the CNN-LSTM model is superior to the single model in such key indicators as Accuracy (94.1%), Recall (92.8%) and F1-score (93.4%), and reduces the false positive rate and false negative rate. It verifies the effectiveness of the CNN-LSTM model in complex attack mode detection and provides a feasible scheme for the optimization of intrusion detection system.