Predictability as a Fine-Grained Measure for Privacy
Researchers have introduced a new privacy framework called "privacy via predictability" that offers a more fine-grained approach than traditional differential privacy (DP). This new method accounts for an attacker's specific knowledge, a compromised portion of the dataset, and the types of queries being made. Predictability measures privacy leakage by assessing how much an attacker's ability to predict sensitive information improves after observing an algorithm's output, beyond what's already known from compromised data. The framework is complementary to DP and can be used alongside it for enhanced privacy control. AI
IMPACT This new privacy framework could lead to more nuanced and effective data protection methods in AI systems.