A Security Concern about Deep Learning Models

摘要：

Content 1.Abstract 2.Background 3.Related work 4.Our work 5.Experiments 6.References 1.Abstract We mainly study on the potential safety hazards in machine learning model by performing an attack on it.In general,the backdoored network can correctly recognize obstacles In special cases,triggering the backdoor in model will change the bounding box and mask

作者: Jiaxi Wu Xiaotong Lin Zhiqiang Lin Yi Tang

作者单位: GUANGZHOU UNIVERSITY

会议类型: 国内会议

会议名称: 第一届网络空间安全科学国际会议(The 1st International Conference on Science of Cyber Security-SciSec 2018)

会议地点: 北京

会议语种:英文

页码: 1-30

在线出版日期: 2018-08-12（万方平台首次上网日期，不代表论文的发表时间）