基于BP网络的入侵检测模型研究与实现

Research of Intrusion Detection Based on Neural Network

入侵检测是一种积极主动的安全防护技术。入侵检测系统可分为基于主机的和基于网络的两种。和防火墙等其它安全产品相比,他们还存在很多缺陷。人工神经网络通过对大量训练样本的学习,可以获得正常和异常数据的分类知识,从而能够对入侵的异常数据进行识别。为此给出了基于BP网络的入侵检测系统,从试验数据发现,该系统不仅在测试阶段的检全率和误检率达到了令人满意的效果,而且在实时检测中,由于计算量不大,对于攻击和扫描的反应速度快,只要建立相应的报警机制,一旦检测到可能的入侵行为,系统就会立即通知管理员采取适当的措施,保护系统安全。

Intrusion Detection is a kind of active voluntary safe protection technique. Intrusion Detection system （IDS） can be divided into Host--based IDS and Network--based IDS. Compared with other safe products such as fire wall, they still have a lot of blemishes. The Artificial Neural Network can acquire the classification knowledge of the normal and abnormal data through a great deal of study on training sample, and carry on identify to the abnormal data for invade. So the author gives the Intrusion Detection System according to BP（Baek Propagation） network. From the experiment data, the system achieves a decent effect on test whole rate and error check rate on testing stage. Moreover, in times of real checking for the quick speed of the reaction in attacking and scaning, as long as setting up the warning mechanism, the system will immediately inform the management to adopt appropriate measure and protect its safety once the possible behavior of invade is examined.