基于机器学习的信号加密有效性识别算法  

Signal Encryption Validity Recognition Algorithm Based on Machine Learning

作　　者：王姗姗 王小青 周军[1] 肖飞[1] WANG Shanshan;WANG Xiaoqing;ZHOU Jun;XIAO Fei(No.30 Institute of CETC,Chengdu Sichuan 610041,China)

机构地区：[1]中国电子科技集团公司第三十研究所,四川成都610041

出　　处：《通信技术》2024年第10期1014-1017,共4页Communications Technology

摘　　要：传统无线信号加密有效性识别算法通常需要对信号进行解调解码,其需要的先验信息多,通用性不强。最近兴起的利用深度学习直接对空口信号进行明密识别的方法解决了传统算法的问题,但对接收信号信噪比要求较高。针对该问题,研究了一种基于机器学习的信号加密有效性识别算法来对空口信号进行眀密检测,该算法不需要掌握被测对象参数,通用性和易用性强,时效性高,并且对接收信号信噪比要求不高,且所需数据量不大,并由于该算法采用接收信号的随机性特征来区分明、密信号,而明、密信号的随机性特征相差很大,因此随着时间的变化其改变很小,解决了现有识别算法的问题。Conventional wireless signal encryption validity recognition algorithms typically require demodulation and decoding of the signal,which requires much prior information,and has low universality.The recently emerged method of using deep learning to directly recognize the clear and secret of air interface signals addresses the issues of conventional algorithms,but it has high demands on the signal-to-noise ratio of the received signals.In this paper,a signal encryption validity recognition approach based on machine learning is investigated to carry out clear and secret detection.The algorithm does not need to possess the parameters of the tested object,demonstrates strong universality,ease of use,high timeliness,has a low requirement for the signal-to-noise ratio of the received signal,and requires a small amount of data.In addition,since the algorithm uses the randomness characteristics of the received signals to distinguish between clear and secret signals,and the randomness characteristics of the clear and secret signals differ significantly,they change minimally with the alteration of time,thus overcoming the drawback of existing recognition methods.

关 键 词：加密有效性识别 特征融合 机器学习 支持向量机 

分 类 号：TN918.9[电子电信—通信与信息系统]