Enlarging hacker's toolbox: deluding image recognition by attacking keypoint orientations - Archive ouverte HAL Access content directly
Conference Papers Year : 2012

Enlarging hacker's toolbox: deluding image recognition by attacking keypoint orientations

Thanh-Toan Do
  • Function : Author
  • PersonId : 870909
Ewa Kijak
Laurent Amsaleg
Teddy Furon

Abstract

Content-Based Image Retrieval Systems (CBIRS) used in forensics related contexts require very good image recognition capabilities. Whereas the robustness criterion has been extensively covered by Computer Vision or Multimedia literature, none of these communities explored the security of CBIR systems. Recently, preliminary studies have shown real systems can be deluded by applying transformations to images that are very specific to the SIFT local description scheme commonly used for recognition. The work presented in this paper adds one %offensive strategy for attacking images, and somehow enlarges the box of tools hackers can use for deluding systems. This paper shows how the orientation of keypoints can be tweaked, which in turn lowers matches since this deeply changes the final SIFT feature vectors. The method learns what visual patch should be applied to change the orientation of keypoints thanks to an SVM-based process. Experiments with a database made of 100,000 real world images confirms the effectiveness of this keypoint-orientation attacking scheme.
Fichier principal
Vignette du fichier
kporientation4.pdf (269.88 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00686809 , version 1 (11-04-2012)

Identifiers

  • HAL Id : hal-00686809 , version 1

Cite

Thanh-Toan Do, Ewa Kijak, Laurent Amsaleg, Teddy Furon. Enlarging hacker's toolbox: deluding image recognition by attacking keypoint orientations. ICASSP - 37th International Conference on Acoustics, Speech, and Signal Processing, Mar 2012, Kyoto, Japan. ⟨hal-00686809⟩
300 View
248 Download

Share

Gmail Facebook X LinkedIn More