User-centered Multiparty Access Control for Collective Content Management
This CAREER project will develop models and techniques to facilitate controlled information sharing of users' data in domains where the data is associated with and co-managed by multiple users, such as bio-repositories, remote teleworking, and social computing. Specific research objectives are: 1) Building on the PI's prior work, develop a foundational model describing access control in terms of the decision making process of a single content manager or content owner, laying the groundwork for the second objective; 2) Develop new models to support synchronous, asynchronous, and combined joint specification of access control policies for shared content for multiple users and site administrators, and 3) Apply those solution concepts to two specific applications, group work and a biobank, and conduct user studies to test goodness of fit, suitability and feasibility of the resulting access setting mechanisms.
This project takes an innovative user-centric approach to ensure that the rigorous models developed result in enforceable mechanisms that can be used on a variety of existing platforms in and multiple domains. To accomplish this, the proposed work draws from multiple disciplines, including access control, game theory for security and privacy, and decision support systems. For example, an individual in a group photo may prefer not to have the photo shared even though others do, but would accept it being shared only with friends; a social network operator wants to maximize use of the system through sharing information while keeping users happy so they remain active. The preferences thus constitute a multi-objective optimization problem. We use a game-theoretic approach to modeling this problem, allowing negotiation to determine access settings. This research will provide users with the ability to express preferred access control settings for shared multi-owned data, jointly influencing with that input the final access settings, while taking into account organizational constraints and existing laws.
Further, we leverage the economic concept of first-mover advantage to create models suitable in scenarios where synchronous coordination among users is not practical. In particular, the models start with extensions and applications of Stackelberg games, and account for the unique constraints occurring with security decisions and uncertainty due to human-bounded rationality. As part of this task, we also develop a method for domain administrators to determine what set of access policy options should be offered to users. Through this administrator-centered model, we enable administrators to identify the set of options that satisfy users' predicted access control decisions and meet their administrator's own internal objectives. We also study hybrid models that build on strengths of synchronous and asynchronous multi-party approaches. The models and mechanisms developed will apply to a wide range of domains, including social computing, remote collaborative content co-authoring, personalized medicine, and genetic data sharing.
The integrated education plan will focus on curriculum enhancement and expanded undergraduate research experiences. We establish a graduate visiting program to give students experience with different disciplines, recognizing the multidisciplinary challenges in information security. The ultimate goal is to produce diverse graduates with the multidisciplinary skills required to design and evaluate IT security solutions.