基于离散序列报文的轮廓格式特征提取方法  

Outline Format Signature Construction Method Based on Separate Protocol Message

作　　者：李阳[1] 李青[1] 张霞[1] LI Yang;LI Qing;ZHANG Xia(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2018年第2期134-139,共6页Journal of Information Engineering University

基　　金：科研基金资助项目(2014500901)

摘　　要：针对格式特征提取算法无法在协议关键字内容不可知的情况下进行特征提取问题,提出一种基于离散序列报文的轮廓格式特征自动提取算法(OSC)。首先对离散序列报文进行二值图像转换使其轮廓格式特征予以显现;接着通过改进的聚类算法将二值图像聚类成簇;最后通过距离加权判决算法提取协议轮廓格式特征进行分类识别。仿真结果表明,OSC对ACARS协议ARINC-618规范中的5种报文类型提取的协议轮廓格式特征与标准格式特征相比,相似度、召回率均达到80%以上。实验结果表明,OSC不依赖完整会话,拥有一定的抗噪能力,更符合实际应用中会话信息不完整及报文内容不可知的情形。To solve the problem that format signature extraction algorithms can＇t be applied when key words are unknowable,a novel outline signature construction method based on separate protocol message（ OSC） is proposed. First,OSC converts separate protocol messages into binary images to make their outline signature prominent. Then OSC puts binary images into clusters with the help of the improved clustering algorithm. Last,OSC acquires the outline signature by distance-weighted decision algorithm in order to classify protocols. Simulation results show that the similarity of five protocols＇ outline signatures extracted by OSC all reach more than 80% compared with the standard pattern. And the recall rate of five protocols also reach more than 80%. Experimental results indicate that the proposed OSC doesn＇t depend on flow and is resistant of noise. OSC is more practical in actual situations.

关 键 词：离散序列报文 协议格式特征 轮廓特征 

分 类 号：TP393.06[自动化与计算机技术—计算机应用技术]