Asgari-Chenaghlu, MeysamFeizi-Derakhshi, Mohammad-RezaNikzad-Khasmakhi, NarjesFeizi-Derakhshi, Ali-RezaRamezani, MajidJahanbakhsh-Nagadeh, ZoleikhaBalafar, Mohammad-AliRahkar-Farshi, Taymaz2021-05-152021-05-1520210020-02551872-6291https://doi.org/10.1016/j.ins.2020.07.007https://hdl.handle.net/20.500.12939/133Rahkar-Farshi, Taymaz/0000-0003-4070-1058; Derakhshi, Mohammad Reza Feizi/0000-0002-8548-976X; Derakhshi, Ali Reza Feizi/0000-0003-3036-1651; Balafar, M.A/0000-0001-5898-0871; Nikzad, Narjes/0000-0003-3536-1343Social media is an inseparable part of our daily life where we post and share photos and media related to our life and in some cases we intend to share them between specific people. This intended and cherry picked sharing of media needs a better solution rather than simply picking users. Some social media platforms do not restrict other users from sharing timeline posts of others; meaning, one can simply forward a post from another person to a third one and no data preservation has been applied. In most cases we do not intend to secure the whole media and only important parts of it are intended to be secured. In this work we propose a novel method based on YoloV3 object detection and chaotic image encryption to overcome the issue of user intended data preservation in social media platforms. Our proposed method is capable of both automatic image encryption on full or user selected regions. Statistical and cryptographic analysis show superiority of our method compared to other state-of-the-art methods while it keeps the speed as high as possible for online and realtime use cases. (C) 2020 Elsevier Inc. All rights reserved.eninfo:eu-repo/semantics/closedAccessDeep LearningChaotic MapsImage EncryptionData PreservationSocial MediaArticle10.1016/j.ins.2020.07.0075422122272-s2.0-85088385948Q1WOS:000574827600012N/A