提出了采用緊致型小波神經(jīng)網(wǎng)絡(luò)來構(gòu)建服務(wù)器預(yù)警系統(tǒng)，將小波和神經(jīng)網(wǎng)絡(luò)直接融合，使網(wǎng)絡(luò)訓練過程從根本上避免了局部最優(yōu)等非線性優(yōu)化問題，小波神經(jīng)元的低相關(guān)性，也使得小波神經(jīng)網(wǎng)絡(luò)有更快的收斂速度；將服務(wù)器中的日志數(shù)據(jù)數(shù)值化后進行網(wǎng)絡(luò)訓練，獲得一個基于小波神經(jīng)網(wǎng)絡(luò)的入侵分類器。實驗結(jié)果，表明小波神經(jīng)網(wǎng)絡(luò)系統(tǒng)自適應(yīng)能力強、學習速度快、預(yù)警精度高、在入侵檢測領(lǐng)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.