Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures
Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures
Matt Fredrikson,S. Jha,Thomas Ristenpart
2015 · DOI: 10.1145/2810103.2813677
Conference on Computer and Communications Security · 引用数 3,201
TLDR
A new class of model inversion attack is developed that exploits confidence values revealed along with predictions and is able to estimate whether a respondent in a lifestyle survey admitted to cheating on their significant other and recover recognizable images of people's faces given only their name.