典型文献
Steganalysis of neural networks based on parameter statistical bias
文献摘要：
Many pretrained deep learning models have been released to help engineers and researchers develop deep learn-ing-based systems or conduct research with minimall effort.Previous work has shown that at secret message can be em-bedded in neural network parameters without compromising the accuracy of the model.Malicious developers can,there-fore,hide malware or other baneful information in pretrained models,causing harm to society.Hence,reliable detection of these vicious pretrained models is urgently needed.We analyze existing approaches for hiding messages and find that they will ineluctably cause biases in the parameter statistics.Therefore,we propose steganalysis methods for stegano-graphy on neural network parameters that extract statistics from benign and malicious models and build classifiers based on the extracted statistics.To the best of our knowledge,this is the first study on neural network steganalysis.The experi-mental results reveal that our proposed algorithm can effectively detect a model with an embedded message.Notably,our detection methods are still valid in cases where the payload of the stego model is low.
文献关键词：
中图分类号：
作者姓名：
Yi Yin;Weiming Zhang;Nenghai Yu;Kejiang Chen
作者机构：
School of Cyber Science and Technology,University of Science and Technology of China,Hefei 230029,China
文献出处：
引用格式：
[1]Yi Yin;Weiming Zhang;Nenghai Yu;Kejiang Chen-.Steganalysis of neural networks based on parameter statistical bias)[J].中国科学技术大学学报,2022(01):1-12
A类：
minimall,baneful,ineluctably,stegano,stego
B类：
Steganalysis,neural,networks,statistical,Many,pretrained,deep,learning,models,have,been,released,help,engineers,researchers,systems,conduct,effort,Previous,has,shown,that,secret,can,parameters,without,compromising,accuracy,Malicious,developers,there,hide,malware,other,information,causing,harm,society,Hence,reliable,detection,these,vicious,urgently,needed,We,analyze,existing,approaches,hiding,messages,find,they,will,cause,biases,statistics,Therefore,we,steganalysis,methods,graphy,from,benign,malicious,build,classifiers,extracted,To,best,our,knowledge,this,first,study,experi,mental,results,reveal,proposed,algorithm,effectively,embedded,Notably,still,valid,cases,where,payload,low
AB值：
0.553769
相似文献
机标中图分类号，由域田数据科技根据网络公开资料自动分析生成，仅供学习研究参考。
