一種新的基于數(shù)據(jù)挖掘技術(shù)的異常入侵檢測(cè)系統(tǒng)研
輸入:數(shù)據(jù)信息E,滑動(dòng)窗口T,時(shí)間t,相似度m,窗口個(gè)數(shù)k,步長(zhǎng)t0,數(shù)據(jù)挖掘規(guī)則庫(kù)K,待測(cè)數(shù)據(jù)規(guī)則為V。
本文引用地址:http://2s4d.com/article/157028.htm
⑤if W={異常}重復(fù)②、③、④ //對(duì)滑動(dòng)時(shí)間窗口得到數(shù)據(jù)規(guī)則集進(jìn)行數(shù)據(jù)評(píng)估;
L=W //每次檢測(cè)結(jié)果提交決策列表以供用戶決策;
5 實(shí)驗(yàn)分析
數(shù)據(jù)參考MIT林肯實(shí)驗(yàn)的DARPA 1999年評(píng)測(cè)數(shù)據(jù)集。由于數(shù)據(jù)信息自身的復(fù)雜性,需要對(duì)數(shù)據(jù)信息進(jìn)行多次訓(xùn)練以降低數(shù)據(jù)噪音的影響。在本實(shí)驗(yàn)中對(duì)ANEIDSDM算法進(jìn)行模擬測(cè)試分為兩個(gè)階段:
(1)為數(shù)據(jù)訓(xùn)練階段:首先收集數(shù)據(jù)信息,依此數(shù)據(jù)信息對(duì)其抽取特征主屬性,挖掘高頻度數(shù)據(jù)項(xiàng)集和低頻數(shù)據(jù)項(xiàng)集,對(duì)高頻數(shù)據(jù)項(xiàng)集進(jìn)行數(shù)據(jù)模式集,對(duì)數(shù)據(jù)模式集進(jìn)行數(shù)據(jù)挖掘,形成數(shù)據(jù)規(guī)則集,最后對(duì)數(shù)據(jù)規(guī)則集進(jìn)行分類,形成標(biāo)準(zhǔn)規(guī)則庫(kù)。實(shí)驗(yàn)時(shí)分為3個(gè)階段收集,實(shí)現(xiàn)3次訓(xùn)練,如表1所示。
(2)數(shù)據(jù)模擬檢測(cè)階段:對(duì)待測(cè)數(shù)據(jù)信息進(jìn)行數(shù)據(jù)規(guī)則集的挖掘,根據(jù)與標(biāo)準(zhǔn)規(guī)則庫(kù)中分類規(guī)則集的相似度對(duì)比,快速分類,通過在線滑動(dòng)窗口和匹配檢測(cè)方法,對(duì)數(shù)據(jù)信息進(jìn)行異常入侵檢測(cè)。若屬于異常信息,則進(jìn)行預(yù)警。實(shí)驗(yàn)時(shí)通過對(duì)7種常見攻擊類型的模式進(jìn)行異常入侵檢測(cè),如表2所示。
通過模擬攻擊實(shí)驗(yàn)表明,數(shù)據(jù)信息經(jīng)過ANEIDSDM入侵檢測(cè)能夠很好地檢測(cè)異常數(shù)據(jù)信息,其誤警率和檢測(cè)率都有了明顯的提高。本實(shí)驗(yàn)同時(shí)可以有效地提高入侵檢測(cè)系統(tǒng)的檢測(cè)速度。
本文針對(duì)現(xiàn)有異常入侵檢測(cè)系統(tǒng)存在的問題,建立了一種新的基于數(shù)據(jù)挖掘技術(shù)的異常入侵檢測(cè)系統(tǒng)模型。該模型包括數(shù)據(jù)采集、數(shù)據(jù)分析、數(shù)據(jù)評(píng)估、事件響應(yīng)等一系列檢測(cè)過程,利用多次訓(xùn)練、滑動(dòng)窗口、規(guī)則分類和相似度匹配思想,大大降低了系統(tǒng)的誤警率,提高了檢測(cè)速度,提升了檢測(cè)率,增強(qiáng)了網(wǎng)絡(luò)系統(tǒng)的安全性能。
參考文獻(xiàn)
[1] VERWORD T,HUNT R. Intrusion detection techniques and approaches[J].Computer Communication,2002,25(15): 1356.1365.
[2] LANE T. Machine learning techniques for the computer security domain of anomaly detection[D]. Purdue University,2000.
[3] MUKKAMALA S, SUNG A H,ABRAHAM A. Intrusion detection using all ensemble of intelligent paradigms[J].Journal of Network and Computer Application,2005,28(2):167-182.
[4] 呂志軍,袁衛(wèi)忠,仲海駿,等. 基于數(shù)據(jù)挖掘的異常入侵檢測(cè)系統(tǒng)研究[J].計(jì)算機(jī)科學(xué),2004,31(10):61-65.
[5] 田新廣,李文法,段σ,等. 基于數(shù)據(jù)挖掘和變長(zhǎng)序列模式匹配的程序行為異常檢測(cè)[J].信號(hào)處理,2008,24(4):521-555.
[6] 凌軍,曹陽(yáng),尹建華,等. 基于時(shí)態(tài)知識(shí)模型的網(wǎng)絡(luò)入侵檢測(cè)方法研究[J].計(jì)算機(jī)學(xué)報(bào),2003,26(11):1591-1597.
[7] 楊向榮,宋擒豹,沈鈞毅,等. 基于數(shù)據(jù)挖掘的智能化入侵檢測(cè)系統(tǒng)[J].計(jì)算機(jī)工程,2001,27(9):17-102.
[8] BARFORD P,HIINE J,PLONKA D,et al. A signal analysis of network traffic anomalies[J].Internet Measurement Workshop,2002,7:1-82.
[9] YE N, LI Xiang Yatig,CHEN Qiang. Probabilistic techniques for intrusion detection based on computer audit data[J]. Man and Cybernetics,Part A,IEEE Transactions on 2001:31(4):266-274.
[10] YE N,EMRAN S M,CHEN Q, et a1. Multivariate statistical analysis of audit trails for host-based intrusion detection[J].IEEE Transactions on Computers,2002,51(7):810-820.
[11] OH S H,LEE W. A clustering based anomaly intrusion detection for a host computer[J].IEICE Transactions on In.formation and Systems,2004,E87-D(8):2086-2094.
[12] HOFMEYR S A,F(xiàn)ORREST S,SOMAYAJI A. Intrusion detection using sequences of system calls[J]. Journal of Computer Security,1998(6):151-180.
[13] LANE T,CARLA E B. An empirical study of two approaches to sequence learning for anomaly detection[J].Machine Learning,2003,51(1):73-107.
評(píng)論