基于社会计算和深度学习的社交网络特定内容监控  

Specific Content Monitoring on Social Networks Based on Social Computing and Deep Learning

作　　者：操晓春[1] 荆丽桦 王蕊[1] 张锐[1] 董振江[2] 熊红凯[3] CAO Xiao-chun JING Li-hua WANG Rui ZHANG Rui DONG Zhen-jiano XIONG Hong-kai(State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China 1 ZTE Corporation,Shenzhen 518057,Chin 2 Department of Electronic Engineering,Shanghai Jiao Tong University,Shanghai,200240,China）3)

机构地区：[1]中国科学院信息工程研究所信息安全国家重点实验室,北京100093 [2]中兴通讯股份有限公司,深圳518057 [3]上海交通大学电子工程系,上海200240

出　　处：《计算机科学》2016年第10期1-8,共8页Computer Science

基　　金：国家重点研发计划(2016YFB0800403);中兴通讯研究基金资助

摘　　要：社交网络极大地方便了人们的生活,加速了信息的共享,但同时也被用于不良和敏感信息的传播,内容安全问题亟待解决。针对此类问题,提出了一套基于社会计算和深度学习的社交网络特定内容监控体系,首先基于成对监督信息实现以内容为导向的半监督社区发现,找到所关心的特定人群;然后对所挖掘的特定人群进行实时监控并获取其发布的内容,对图像和视频进行实时自动内容识别;同时针对实网数据误报多的问题提出面向多负类的误判修正方法,以达到收集实时信息,净化网络环境,在一定程度上预防犯罪的目的。Social networks provide great convenience to people＇s daily life and information sharing. Unfortunately, these conveniences are accompanied with content security problems, where the social networks are frequently employed to dis- seminate malicious or sensitive information. This paper proposed a content security solution, which builds a system to monitor specific content based on social computing and deep learning. To search for the specific people, the system aehieves a content-sensitive semi-supervised community discovery method with pairwise constraint. By monitoring the discovered people and obtaining their published content, the system performs an automatic detection procedure to identi- fy the content of the published images and videos. In addition, an error correction method was proposed to reduce the false positives when processing the real network data. Experimental results demonstrate that the proposed system gives decent performances under various circumstances.

关 键 词：社会计算 社区发现 深度学习 图像识别 社交网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]