An EU team aimed to develop software error-removal tools specifically for differential privacy uses. The developments address resource consumption and privacy budget, with the prototype algorithm showing improved performance.
The social costs of defective software can be prevented through removing design errors, and modern programming languages include error-removal tools. Generally, however, such techniques do not consider resource consumption.
With EU funding, the PLATFORM (Practical light types for resource consumption) project aimed to develop software error-removal tools addressing resource consumption. The efforts were focused on differential privacy applications. The two-partner project ran for three years from mid-2011.
First-year work involved study of a type-based framework, useful in inferring an application's execution time consumption. The benefit was the ability to analyse different resolution models, making the technique widely applicable and modular.
During the following year, the researchers designed a methodology, using a domain-specific programming language, for automatically determining the privacy budget used by an application.
The final year's work focused on applying the previously developed techniques to both resource consumption and differential privacy. Researchers extended the theory and practice of the latter, including an economic model describing how to set the relevant defining parameters. The resulting algorithm showed good performance. The project's novel verification technique also proved useful.
PLATFORM achieved a new theory applicable to uniformly describing several resource analyses. An additional result was researcher training and development.