Probabilistic modeling goal of DARPA’s Brandeis program
CAMBRIDGE, Mass. Charles River Analytics Inc., developer of intelligent systems solutions, partnered with Galois, Inc. to integrate probabilistic modeling into the Defense Advanced Research Projects Agency’s (DARPA) Brandeis program.
Under the Brandeis program — which seeks to develop the technical means to protect the private and proprietary information of individuals and enterprises — the Galois-led TAMBA team developed technology to evaluate the effectiveness of privacy-aware systems.
Some new privacy systems employ encryption and differential privacy algorithms, with the goal of preventing others from inferring sensitive data. Charles River Analytics developed an approach to test these systems—inference algorithms for probabilistic models are used to determine whether the sensitive information under protection can be inferred.
The approach incorporates side-channel information and complex domain prior models to evaluate the strength of the system against sophisticated adversaries. This probabilistic approach also provided a platform for improving the utility of these systems by enhancing accuracy without reducing privacy, according to the company.