本文選題：神經(jīng)網(wǎng)絡(luò) + 入侵檢測��； 參考：《計算機(jī)仿真》2013年09期

【摘要】：研究網(wǎng)絡(luò)節(jié)點(diǎn)危險程度評估優(yōu)化入侵檢測問題。由于入侵的多樣性和隨機(jī)性,造成準(zhǔn)確檢測困難。傳統(tǒng)的網(wǎng)絡(luò)安全模型都是對信譽(yù)度或信任度等概念完成惡意節(jié)點(diǎn)整體檢測,因?yàn)閱蝹�節(jié)點(diǎn)屬性較為復(fù)雜,所承擔(dān)的作用不同,使得針對單個節(jié)點(diǎn)信息評估過程較為粗糙,很難設(shè)定準(zhǔn)確閥值進(jìn)行精確判斷,造成傳統(tǒng)模型對單個節(jié)點(diǎn)危險程度評估不準(zhǔn)。提出一種危險程度的網(wǎng)絡(luò)節(jié)點(diǎn)惡意程度評估模型,使用馬爾科夫算法與貝葉斯學(xué)習(xí)器計算單個節(jié)點(diǎn)的危險度,運(yùn)用貝葉斯方法推斷出節(jié)點(diǎn)惡意程度的解空間,依據(jù)節(jié)點(diǎn)的屬性特征計算節(jié)點(diǎn)的惡意度,克服傳統(tǒng)方法不能對單個節(jié)點(diǎn)做出判斷的弊端。實(shí)驗(yàn)表明,與已有的安全模型相比,提出的安全管理模型對惡意節(jié)點(diǎn)具有更高的檢測率。
[Abstract]:Research on network node risk assessment and optimization of intrusion detection problem. Due to the diversity and randomness of intrusion, it causes accurate detection difficulties. The evaluation process of the node information is relatively rough, it is difficult to set accurate thresholds for accurate judgment, which causes the traditional model to evaluate the risk degree of single node. A risk degree evaluation model of network node malware is proposed, and the Markoff algorithm and Bayesian classifier are used to calculate the risk degree of a single node, and the Bayesian formula is used. The method deduce the solution space of the node's malicious degree, calculate the malicious degree of the node according to the attribute characteristics of the node, overcome the disadvantage that the traditional method can not judge the single node. The experiment shows that the proposed security management model has a higher detection rate to the malicious node compared with the existing security model.
【作者單位】： 貴州大學(xué) 計算機(jī)科學(xué)與信息學(xué)院;
【分類號】：TP393.08

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 穆成坡;黃厚寬;田盛豐;;入侵檢測系統(tǒng)報警信息聚合與關(guān)聯(lián)技術(shù)研究綜述[J];計算機(jī)研究與發(fā)展;2006年01期

2 周曄,楊天奇;一種基于置信度的異常檢測模型與設(shè)計[J];計算機(jī)仿真;2005年01期

3 肖海軍;洪帆;張昭理;廖俊國;;基于融合分類和支持向量機(jī)的入侵檢測研究[J];計算機(jī)仿真;2008年04期

4 周荃;王崇駿;王s，

本文編號：1902418