Graduate School and Research Center in Digital Sciences

Automatic face anonymization in visual data: Are we really well protected?

Ruchaud, Natacha; Dugelay, Jean-Luc

EI 2016, ISAT International Symposium on Electronic Imaging, February 14-18, 2016, San Francisco, USA

With the proliferation of digital visual data in diverse domains (video surveillance, social networks, medias, etc.), privacy concerns increase. Obscuring faces in images and videos is one option to preserve privacy while keeping a certain level of quality and intelligibility of the video. Most popular filters are blackener (black masking), pixelization and blurring. Even if it appears efficient at first sight, in terms of human perception, we demonstrate in this article that as soon as the category and the strength of the filter used to obscure faces can be (automatically) identified, there exist in the literature ad-hoc powerful approaches enable to partially cancel the impact of such filters with regards to automatic face recognition. Hence, evaluation is expressed in terms of face recognition rate associated with clean, obscured and de-obscured face images.

Document Hal Bibtex

Title:Automatic face anonymization in visual data: Are we really well protected?
City:San Francisco
Department:Digital Security
Eurecom ref:4914
Copyright: IS&T
Bibtex: @inproceedings{EURECOM+4914, year = {2016}, title = {{A}utomatic face anonymization in visual data: {A}re we really well protected?}, author = {{R}uchaud, {N}atacha and {D}ugelay, {J}ean-{L}uc}, booktitle = {{EI} 2016, {ISAT} {I}nternational {S}ymposium on {E}lectronic {I}maging, {F}ebruary 14-18, 2016, {S}an {F}rancisco, {USA}}, address = {{S}an {F}rancisco, {UNITED} {STATES}}, month = {02}, url = {} }
See also: