针对音视频统一信息服务系统发布异常内容的安全性增强研究

Research on Security Enhancement for Abnormal Content Release in Unified Audio Video Information service Systems

[目的]传统的人工审核已无法满足信息实时性和准确性的要求。为增强上海轨道交通的安全性,有必要对音视频统一信息服务系统发布的异常内容进行智能审核。[方法]将运用决策树算法的内容审计系统与音视频统一信息服务系统相结合,构建集成数据与多维信息安全模型,以此完善既有的音视频信息安全管理系统,对音视频发布内容进行监测和审核,以识别并应对潜在的安全威胁。通过试验对音视频信息安全管理系统在不同工况的应用适用性及异常审核性能进行评估和验证。[结果及结论]完善后的音视频信息安全管理系统具有充分的实用价值,其漏检率和误检率分别控制在0.5%以内和2.0%以内,既能高效检测违规内容,又能保证正常内容不受干扰;服务稳定性结果为99.9%,确保长时间连续监控;视频流送检延时不超过5 s,异常样本检出到告警时间不超过2 s,可快速启动应对措施。

[Objective]Traditional manual review processes can no longer meet the demands for real-time and accurate information assessment.To enhance the security of Shanghai rail transit system,it is necessary to implement an intelligent review mechanism for identifying abnormal content within unified audio video information service systems.[Method]The content audit system using decision tree algorithms is combined with a unified audio video information service system,to construct an integrated data and multi-dimensional information security model,thereby improving the existing audio video information security management system.This system monitors and audits audio video content to identify and address potential security threats.Experimental evaluations are conducted to assess the system applicability and the performance of abnormal content review under various operational conditions of the audio video information security management system.[Result&Conclusion]The improved audio video information security management system demonstrates significant practical value,with both omission ratio and error detection rate controlled within 0.5%and 2.0%,respectively.The system efficiently detects prohibited content while ensuring that legitimate content remains unaffected.The service stability is recorded at 99.9%,ensuring continuous monitoring over long periods.Video stream inspection latency is controlled within 5 seconds,and the time from abnormal sample detection to alert within 2 seconds,allowing for rapid response measures.