On the vulnerability of deep learning to adversarial attacks for camera model identification
作者：
Highlights：
•
• This paper studies the vulnerability of deep learning approaches for camera model identification.
• Adversarial examples are generated in a realistic scenario where the image undergoes a lossless or lossy compression.
• Adversarial training is carried out to increase the robustness of CNN-based detectors.
• It is investigated the transferability of attacks among different networks.
摘要
•This paper studies the vulnerability of deep learning approaches for camera model identification.•Adversarial examples are generated in a realistic scenario where the image undergoes a lossless or lossy compression.•Adversarial training is carried out to increase the robustness of CNN-based detectors.•It is investigated the transferability of attacks among different networks.
论文关键词：Camera model identification,Convolutional neural networks,Adversarial attacks
论文评审过程：Received 11 December 2017, Revised 24 February 2018, Accepted 6 April 2018, Available online 13 April 2018, Version of Record 7 May 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.04.007
