Â鶹´«Ã½AV

Protect-IT targets to attract studies on security and privacy for machine learning (ML) from an information-theoretic standpoint. Accuracy and efficiency of ML systems are ensured by employing large datasets which usually contain highly sensitive/personal information. This strong dependence on personal information jeopardizes the privacy and security of innocent Internet users who are contributing, knowingly or not, to these online statistical datasets. Protect-IT aims to bring the typical attendees of CSF, who have expertise on the theory of cryptography and (algorithmic) fairness together with researchers on information theory to study, develop, and evaluate privacy, security, and fairness attacks against ML along with defense strategies to counter them. We put information theory at the heart of this endeavor and call for contributions grounded in information-theoretic concepts and principles, aiming to enrich preliminary research efforts and to achieve widespread adoption. Topics of interests include but not limited to:

- Information leakage, data correlation 
- Defining and quantifying privacy
- Differential privacy and private data analysis
- Cryptographic tools for privacy
- Machine learning and privacy
- Transparency, robustness and abuse in privacy systems
- Information theoretic foundations of security and privacy
- Algorithmic fairness 
- Adversarial learning 
- Intrusion detection
- Abstract attacks on privacy
- Security attacks and defenses