一個以快速粒子群最佳化為基礎的高效能入侵偵測系統

CNU IR > College of Sustainable Environment and Recreation > Dept. of Applied Geoinformatics > MOST Project > Item 310902800/24940

Please use this identifier to cite or link to this item: https://ir.cnu.edu.tw/handle/310902800/24940

Title:	一個以快速粒子群最佳化為基礎的高效能入侵偵測系統 An Efficient Intrusion Detection System Based on Fast Particle Swarm Optimization
Authors:	蔡崇煒楊竹星
Contributors:	應用空間資訊系
Date:	2011
Issue Date:	2012-01-03 13:57:18 (UTC+8)
Abstract:	本計畫為兩年期的研究計畫，主要目的在建置一個快速的智慧型入侵偵測系統 (fast intelligent intrusion detection system; FIIDS)，以解決傳統分群演算法應用至入侵偵測系統，需面臨兩項重要問題： 1、『花費巨量的計算資源』，以及2、當新資料進入系統時，需『重新執行分群演算法』。為了解決上述兩項重要的研究議題，本計畫將設計一個可自我學習的啟發式演算法 (metaheuristics)，分析及過濾不當資訊，以提供更有效的「網路應用之判別與控管系統」服務。本計畫第一年，將嘗試在不影響判別正確率，或犧牲小幅正確判別的情況下，加速入侵偵測系統的速度。藉由去除重複計算，設計一個快速分群演算法 (fast clustering algorithm; FCA)，並整合現有的資訊檢索方法，分析這些輸入資訊的相似度、關連性等屬性，以提升演算法正確率。此外，本計畫擬在第一年完成入侵偵測系統之雛形，以便於測試演算法效能及系統成效。本計畫第二年，目標在不影響或小幅影響判別正確率的情況下，自動更新入侵偵測系統的分類器，使系統有能力在面臨新型或變種的網路攻擊行為時，快速偵測出這些不當的網路行為，並加以管控。換言之，延續第一年的成果，我們將設計一個快速漸進式分群演算法 (fast incremental clustering algorithm; FICA)。藉由漸進式的自主學習，以提升系統對於流量分析的正確性。這項演算法將可隨著新的資料型態進入系統，動態更新分類規則。另外，我們將利用所設計之演算法建置網路應用的判別與控管服務，並實作封包擷取、與快速判別的技術提供網路流量分類。本計畫將完成一個快速且可自主學習的入侵偵測系統。這項系統將具能力判別新型態的不當網路行為，以幫助網路管理人員分析及預防日新月異的不當網路行為。 In this two-year project, a fast intelligent intrusion detection system, called FIIDS, will be designed and implemented to solve two of the major problems in application to intrusion detection system. The problems are: (1) “a great deal of computation resource is required,” and “the traditional clustering algorithm has to be performed again from scratch when facing new network application types.” By using a self-learning metaheuristic algorithm to analyze and filter inappropriate information, this system can provide a more efficient service to recognize and manage the network applications. In the first year, a fast clustering algorithm (FCA) will be designed and developed. By eliminating the redundant computations of traditional clustering algorithms, integrating the information retrieval technologies, and analyzing the similarities, relationships, and attributes of the input data, the accuracy rate can be increased. A prototype system will also be developed and implemented to test and measure the performance of the system. In the second year, a fast incremental clustering algorithm (FICA) will be designed and developed. Because incremental clustering algorithm can be used to dynamically update the classification rules when new data are detected, the accuracy rate can be further improved. Then comes application of this new algorithm to intrusion detection system (including packet extraction, detection) to detect and manage the network applications. In summary, this project is aimed at developing a fast and self-learning intrusion detection system. The proposed system can recognize new types of irregular network behaviors to help the network manager detect, analyze, and prevent those behaviors more effectively and efficiently.
Relation:	計畫編號：100-2218-E-041-001-MY2 計畫年度：100；起迄日期：2011-01-01~2012-07-31 核定金額：447,000元
Appears in Collections:	[Dept. of Applied Geoinformatics] MOST Project

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	2868	View/Open

Loading...