提出了采用緊致型小波神經網絡來構建服務器預警系統，將小波和神經網絡直接融合，使網絡訓練過程從根本上避免了局部最優等非線性優化問題，小波神經元的低相關性，也使得小波神經網絡有更快的收斂速度；將服務器中的日志數據數值化后進行網絡訓練，獲得一個基于小波神經網絡的入侵分類器。實驗結果，表明小波神經網絡系統自適應能力強、學習速度快、預警精度高、在入侵檢測領域有良好的實用性。A server warning system based on wavelet neural network is presented. Wavelet neural network is a kind of neural networks, which combines wavelet theory with neural network theory, that is wavelet function forming neuron, it avoid the problem of nonlinear optimizations, such as local optimization. WNN has fast convergence because of the low correlation of wavelet neuron. In this system ,the WNN was trained by using numbered records of system log files in server and the classifier for intrusion detection can be obtained. The experimental result shows that the performance of this system is highly adaptive , the learning speed is fast, and the rate of warning accurate is high, so it is practicable in intrusion detection.