With the rapid developments of Internet of Things (IoT) and proliferation of embedded de-vices, large volume of personal data are collected, which however, might carry massive private informa-tion about attributes that users do not want to share. Many privacy-preserving methods have been proposed to prevent privacy leakage by perturbing raw data or extracting task-oriented features at local devices. Un-fortunately, they would suffer from significant pri-vacy leakage and accuracy drop when applied to other tasks as they are designed and optimized for prede-fined tasks. In this paper, we propose a novel task-free privacy-preserving data collection method via adver-sarial representation learning, called TF-ARL, to pro-tect private attributes specified by users while main-taining data utility for unknown downstream tasks. To this end, we first propose a privacy adversarial learning mechanism (PAL) to protect private attributes by op-timizing the feature extractor to maximize the adver-sary's prediction uncertainty on private attributes, and then design a conditional decoding mechanism (Con-Dec) to maintain data utility for downstream tasks by minimizing the conditional reconstruction error from the sanitized features. With the joint learning of PAL and ConDec, we can learn a privacy-aware feature ex-tractor where the sanitized features maintain the dis-criminative information except privacy. Extensive ex-perimental results on real-world datasets demonstrate the effectiveness of TF-ARL.

School of Cyber Science and Engineering,Wuhan University,430072 China;School of Cyber Science and Technology,Zhejiang University,310027 China;State Key Laboratory of Integrated Services Networks,Xidian University,710071 China;Department of Computer Science,College of William and Mary,Williamsburg,VA,23186 USA

[1]Zhibo Wang;Wei Yuan;Xiaoyi Pang;Jingxin Li;Huajie Shao-.Towards Task-Free Privacy-Preserving Data Collection)[J].中国通信（英文版）,2022(07):310-323

prede,sanitized,ConDec,criminative

Towards,Task,Free,Privacy,Preserving,Data,Collection,With,rapid,developments,Internet,Things,IoT,proliferation,embedded,large,volume,personal,collected,which,however,might,carry,massive,private,about,attributes,that,users,not,want,share,Many,privacy,preserving,methods,have,been,proposed,prevent,leakage,by,perturbing,raw,extracting,oriented,features,local,devices,Un,fortunately,they,would,suffer,from,significant,accuracy,drop,when,applied,other,tasks,designed,optimized,fined,this,paper,novel,free,collection,via,representation,learning,called,TF,ARL,specified,while,taining,utility,unknown,downstream,end,first,adversarial,mechanism,PAL,protect,timizing,extractor,maximize,sary,prediction,uncertainty,then,conditional,decoding,maintain,minimizing,reconstruction,error,joint,aware,where,dis,information,except,Extensive,perimental,results,real,world,datasets,demonstrate,effectiveness

0.52664