科技通报
科技通報
과기통보
BULLETIN OF SCIENCE AND TECHNOLOGY
2015年
6期
109-111
,共3页
近场源%网络攻击%盲源分离
近場源%網絡攻擊%盲源分離
근장원%망락공격%맹원분리
near field source%network attack%blind source separation
多目标无限方差网络攻击作为一种新的病毒攻击方式，对网络安全造成严重威胁，通过对多目标攻击源的时频盲源分离，准确检测和发掘攻击信号源和特征，提高检测性能。传统方法中对攻击源的分离算法采用随机场谱峰搜索算法，存在计算量大的问题。提出一种基于近场源DOA估计的多目标无限方差网络攻击下时频盲源分离算法。为了提高多目标无限方差网络攻击近场源的识别概率，对提取得到的波束域约束指向性特征输出结果进行频分复用分解，通过时频特征提取，得到相关的攻击信号源参数，在多目标无限方差网络攻击近场源模型中，设计匹配滤波器，把近场源的参数估计问题变成了一个三维参数联合DOA估计问题，实现对网络攻击源时频盲源分离。仿真结果表明，具有较好的网络攻击时频盲分离性能，对攻击源估计的均方根误差较传统算法明显减小，实现了对攻击信源的准确识别和检测。
多目標無限方差網絡攻擊作為一種新的病毒攻擊方式，對網絡安全造成嚴重威脅，通過對多目標攻擊源的時頻盲源分離，準確檢測和髮掘攻擊信號源和特徵，提高檢測性能。傳統方法中對攻擊源的分離算法採用隨機場譜峰搜索算法，存在計算量大的問題。提齣一種基于近場源DOA估計的多目標無限方差網絡攻擊下時頻盲源分離算法。為瞭提高多目標無限方差網絡攻擊近場源的識彆概率，對提取得到的波束域約束指嚮性特徵輸齣結果進行頻分複用分解，通過時頻特徵提取，得到相關的攻擊信號源參數，在多目標無限方差網絡攻擊近場源模型中，設計匹配濾波器，把近場源的參數估計問題變成瞭一箇三維參數聯閤DOA估計問題，實現對網絡攻擊源時頻盲源分離。倣真結果錶明，具有較好的網絡攻擊時頻盲分離性能，對攻擊源估計的均方根誤差較傳統算法明顯減小，實現瞭對攻擊信源的準確識彆和檢測。
다목표무한방차망락공격작위일충신적병독공격방식，대망락안전조성엄중위협，통과대다목표공격원적시빈맹원분리，준학검측화발굴공격신호원화특정，제고검측성능。전통방법중대공격원적분리산법채용수궤장보봉수색산법，존재계산량대적문제。제출일충기우근장원DOA고계적다목표무한방차망락공격하시빈맹원분리산법。위료제고다목표무한방차망락공격근장원적식별개솔，대제취득도적파속역약속지향성특정수출결과진행빈분복용분해，통과시빈특정제취，득도상관적공격신호원삼수，재다목표무한방차망락공격근장원모형중，설계필배려파기，파근장원적삼수고계문제변성료일개삼유삼수연합DOA고계문제，실현대망락공격원시빈맹원분리。방진결과표명，구유교호적망락공격시빈맹분리성능，대공격원고계적균방근오차교전통산법명현감소，실현료대공격신원적준학식별화검측。
Multi objective infinite variance cyber attacks as a new kind of virus attacks, poses a serious threat to network se?curity, separated by time frequency blind source of multi target attack source, accurate detection and explore the attack sig?nal source and characteristics, to improve the detection performance. In the traditional method of attack source separation algorithm using random field spectral peak searching algorithm, the problems of large amount of calculation of the problem. A method based on multi objective infinite variance of network attack near field source DOA estimation of time-frequency blind source separation algorithm is proposed. In order to improve target infinite variance of network attack near field source identification probability of beam domain constraints extracted points to the result output characteristics of frequen?cy division multiplexing frequency decomposition, feature extraction through, get the attack signal source parameters, in the near field source model of multi objectives and infinite variance of network attack, the design of matched filter the near field source parameter estimation problem into a three-dimensional parameter estimation problem combined with DOA, the realization of frequency blind source separation of network attack source. The simulation results show that the network has better attack frequency blind separation performance, the attack source estimation of root mean square error decreases obvi?ously compared to the traditional algorithm, the realization of accurate identification and detection of the attack source.
