應用類神經網路於阻斷式服務攻擊之預測

Forecast of Denial of Service (DoS) Attacks via Neural Networks

紀宏宜 ； 張偉德 ； 陳志榮

《網際網路技術學刊》 9卷2期 (2008/03) Pp. 173-178

https://doi.org/10.6138/JIT.2008.9.2.10

摘要

本文的目的是利用倒傳遞類神經網路來偵測伺服主機是否遭受阻斷服務攻擊。如我們所知，類神經網路是模擬生物神經傳遞訊息之功能，具有學習與記憶之能力，應用層面相當廣泛，而倒傳遞類神經網路模式是目前神經網路中最具代表性之一，屬於監督式學習網路，適合於系統診斷、預測等應用。此外阻斷服務攻擊是目前網際網路應用上，令管理人員備感困擾之網路安全問題，其攻擊特性為針對某一網路服務，湧入大量網路訊息要求服務，使得提供服務之網路伺服主機，陷入忙碌運作的陷阱中，無法提供其他使用者正常服務。由於該攻擊方式尚未出現有效之預防措施，因此本文嘗試透過倒傳遞類神經網路，事先學習主機在異常以及正常的運作狀況下之訓練資料，以期進一步應用在實際網路攻擊之預測。

關鍵字

阻斷服務攻擊 ； 網路安全 ； 類神經網路

並列摘要

The forecast of Denial of Service (DoS) attacks is achieved via the back-propagation neural networks in this paper. It is well known that artificial neural network is to simulate the behavior of real biological neuron and has successfully applied to a variety of application fields due to its great learning capabilities. The back-propagation neural network belongs to the category of supervisory networks, which is suitably used for system diagnoses and forecasts. Moreover, the DoS attack is a quite significant topic on the network security and always bores the network managers. The main feature of the DoS attack is that a large number of clients send the service requests simultaneously to certain server via the internet such that this server is too busy to provide normal services for others. Until now, there is no good and effective solution for solving this kind of attack problem. Therefore, in this paper the back-propagation neural network is used to learn the training data generated from the normal and abnormal operations of the server, and further to forecast whether the server is attacked by the DoS.

並列關鍵字

Denial of Service DoS ； Network Security ； Neural Networks