Machine learning and data mining algorithms have penetrated our everyday lives and play a central role in many application domains, including social networks, healthcare, location-based systems, and advertising. At the same time, 90% of today's data have been produced in the last two years. These data come from social networking sites, mobile phone applications, electronic medical record systems, e-commerce sites, and open data portals. The analysis of this wealth of data can lead to valuable insights that will benefit data recipients and the society at large, but may lead to serious privacy breaches, unless privacy-enhancing technologies are in place.

To address the increasing privacy concerns of both individuals and organizations about the use of their data, the domain of privacy-preserving data mining was brought into existence, leading to a wide range of methods proposed by industry and academia, such as the development of differential privacy and utility-preserving anonymity. However, emerging applications pose new challenges for privacy-preserving methods, which are related to the type and intended use of the data, while privacy should be embedded throughout the lifecycle of the data, from data collection to data sharing and use. In fact, there are several interesting issues that require further investigation from both a theoretical and a practical perspective.

First, there is a need to address the challenge of preserving privacy in emerging applications, such as social networks, healthcare, cloud computing, e-commerce and location-based systems. Protecting data in these applications is particularly challenging, because they involve large volumes of complex, heterogeneous, and dynamic data. These data must be protected from several attacks that may lead to disclosure of private or sensitive information and discrimination, using scalable methods.

Second, ensuring that privacy-protected data remain useful in intended applications, such as building accurate data mining models or enabling complex analytic tasks, is essential. This calls for a careful study of utility requirements, and the development of algorithms to enforce these requirements while protecting data privacy. Meanwhile, benchmarks for evaluating data utility and large-scale case studies, which will enable the reliable and meaningful comparison of competing methods, need to be developed.

Third, privacy-preserving data integration and linkage still remain very challenging tasks that call for new approaches. This is because, there is a large number of data providers and recipients, with different and possibly competing privacy and utility requirements, who need to co-operate to produce privacy-protected data or collaborate to perform a data mining task in a privacy-preserving way.

This workshop aims at encouraging principled privacy research that develops methodologies to address the above issues. Topics of this workshop include novel methods to address modern research challenges, such as:

How to guard against attackers who may manipulate the data prior to anonymization, exploit knowledge of the privacy mechanism, combine multiple datasets together, collude with other attackers, or employ data mining algorithms to breach privacy?
What are good design guidelines for privacy definitions?
How can we compare the utility offered by different privacy definitions and methods?
How can we develop models and benchmarks to evaluate data utility, across different privacy definitions?
What are information-theoretic or computational barriers to privacy and utility?
What new privacy challenges are posed by emerging applications, such as location-based social networks and large-scale healthcare systems, and how can they be addressed?


PADM will be held in conjunction with the IEEE ICDM 2013 conference in Dallas, Texas. The purpose of this workshop is to encourage principled research that develops methodologies to address open privacy problems.