Researchadversarial attacksface recognitionprivacy tools
University of Toronto Engineers Disable Facial Recognition
8.6
University of Toronto engineering researchers led by Parham Aarabi and Avishek Bose recently developed an adversarial neural-network filter that disrupts automated facial recognition. Tested on the 300-W dataset of over 600 faces, the method reduced detectable faces from 100% to 0.5%, altering pixels imperceptibly to humans while foiling detectors. The team plans to make the privacy filter available as a mobile or web app.
Key Points
- 1Develop paired neural networks that generate adversarial perturbations, reducing detectable faces from 100% to 0.5%
- 2Expose critical detector vulnerabilities by targeting facial landmarks with near-imperceptible pixel-level attacks
- 3Enable practitioners to build privacy tools (apps/web) preventing automated face recognition and attribute extraction
Scoring Rationale
Strong experimental results and clear practical utility, limited by moderate novelty and need for broader validation.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
Active Search Campaigns by BudgetEasyHigh CPC Clicks & Poor Landing PagesMediumCampaign ROAS by Attribution ModelHard
250 free problems · No credit card
See all Ad Tech problems
