PI: Vitaly Shmatikov (The University of Texas at Austin) Co-PI: Joan Feigenbaum (Yale University)
When collaborations involve sharing of sensitive data about individuals and organizations, these data must be protected from unauthorized searches, abuse, and misuse. Conventional interpretations of privacy as confidentiality and/or inaccessibility of any individual piece of information are inadequate in collaborative environments, where some collaborators may legitimately be allowed to access parts of the joint dataset, and no trust assumptions can be made about their computing platforms.
The goal of this project is to develop new concepts and frameworks for privacy in collaborative environments, focusing on global properties of the joint dataset such as security against unreasonable searches and abusive information harvesting. Techniques include provably secure data transformations that assure global and individual privacy properties after information has been released in response to a legitimate request. Another research objective is a theory of privacy that explicitly incorporates economic measures of information value. To enforce global privacy policies, this project will develop new cryptographic techniques for dataset obfuscation and sanitization, ensuring that only policy-compliant queries can be computed on the dataset after it has been transferred to the collaborators.
The main objective is to design privacy-preserving data transformations that are provably secure without unrealistic assumptions about "tamper-proof" software or hardware. Privacy technologies developed in the course of the project will enable important collaborative applications, ranging from joint analysis of patient data in multi-institution clinical trials to transaction monitoring by law enforcement agencies that complies with the citizens' Fourth Amendment right to be secure against unreasonable searches.
Project URL: www.cs.utexas.edu/~shmat/privacyframeworks/
