Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples

Abstract: 

We investigate if the random feature selection approach proposed in [1] to improve the robustness of forensic detectors to targeted attacks, can be extended to detectors based on deep learning features. In particular, we study the transferability of adversarial examples targeting an original CNN image manipulation detector to other detectors (a fully connected neural network and a linear SVM) that rely on a random subset of the features extracted from the flatten layer of the original network. The results we got by considering three image manipulation detection tasks (resizing, median filtering and adaptive histogram equalization), two original network architectures and three classes of attacks show that feature randomization helps to hinder attack transferability, even if, in some cases, simply changing the architecture of the detector, or even retraining the detector is enough to prevent the transferability of the attacks.

up
0 users have voted:

Paper Details

Authors:
Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang
Submitted On:
30 January 2020 - 12:03pm
Short Link:
Type:
Research Manuscript
Event:
Presenter's Name:
Mauro Barni, Ehsan Nowroozi, Benedetta Todi, Bowen Zhang
Paper Code:
3729
Document Year:
2020
Cite

Document Files

ICASSP20-final..pdf

(34)

Subscribe

[1] Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang, "Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples", IEEE SigPort, 2020. [Online]. Available: http://sigport.org/4967. Accessed: Jun. 04, 2020.
@article{4967-20,
url = {http://sigport.org/4967},
author = {Mauro Barni; Ehsan Nowroozi; Benedetta Tondi; Bowen Zhang },
publisher = {IEEE SigPort},
title = {Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples},
year = {2020} }
TY - EJOUR
T1 - Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples
AU - Mauro Barni; Ehsan Nowroozi; Benedetta Tondi; Bowen Zhang
PY - 2020
PB - IEEE SigPort
UR - http://sigport.org/4967
ER -
Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang. (2020). Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples. IEEE SigPort. http://sigport.org/4967
Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang, 2020. Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples. Available at: http://sigport.org/4967.
Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang. (2020). "Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples." Web.
1. Mauro Barni, Ehsan Nowroozi, Benedetta Tondi, Bowen Zhang. Effectiveness of Random Deep Feature Selection for securing image manipulation detectors against adversarial examples [Internet]. IEEE SigPort; 2020. Available from : http://sigport.org/4967