基于深度学习的非结构化敏感信息识别系统设计  

Design of Unstructured Sensitive Information Recognition System Based on Deep Learning

作　　者：庞宁 PANG Ning(Shenzhen Power Supply Co.,Ltd.,Shenzhen 518048,China)

机构地区：[1]深圳供电局有限公司,广东深圳518048

出　　处：《微型电脑应用》2023年第7期146-148,共3页Microcomputer Applications

摘　　要：由于当前的敏感信息识别系统未考虑建立特定的敏感字仓库,容易导致不同文本中敏感信息识别精度低、识别效果差、漏报和误报数量大的问题,为此,设计了基于深度学习的非结构化敏感信息识别系统。这个系统采用以低功耗高通信速度为主要功能的DM9000网卡芯片,完成数据传送与接收工作,选取基于TMS320F28335的超声流量器,并应用于与网卡芯片连接的扩展网络端口部分,完成系统硬件设计。计算非结构化敏感信息语义结构和敏感词出现概率,利用深度学习建立敏感字仓库,设置真假判定函数,获取生成网络的最大值与最小值,得到权值属性维度,生成敏感信息识别算法,计算损失函数极小值,实现非结构化敏感信息识别。实验结果表明,在单字识别、词汇识别、短语识别、长句识别中,所提方法的敏感信息识别效果较好,识别漏报和误报数量较小,能够有效提高不同文本中敏感信息识别精度。In view of the problems that the current sensitive information recognition system does not consider to establish a specific sensitive word warehouse,resulting in low recognition accuracy,poor recognition effect and large number of missing and false positives of sensitive information in different texts,an unstructured sensitive information recognition system based on deep learning is designed.The DM9000 network card chip with low power consumption and high communication speed is used to complete the data transmission and reception.The ultrasonic flowmeter based on TMS320F28335 is selected and applied to the extended network port connected with the network card chip to complete the system hardware design.It calculates the semantic structure of unstructured sensitive information and the occurrence probability of sensitive words,uses deep learning to establish a sensitive word warehouse,sets the true and false judgment function,obtains the maximum and minimum values of the generated network,obtains the weight attribute dimension,generates the sensitive information identification algorithm,calculates the minimum value of the loss function,and realizes the unstructured sensitive information identification.The experimental results show that in single word recognition,vocabulary recognition,phrase recognition and long sentence recognition,the proposed method has better sensitive information recognition effect,and the number of missing and false positives is small,which can effectively improve the recognition accuracy of sensitive information in different texts.

关 键 词：深度学习 非结构化系统 敏感信息 识别系统 信息识别 网卡芯片 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]