Stronger Privacy Preserving Projections for Multi-Agent Planning. Shani, G., Maliah, S., & Stern, R. In
Stronger Privacy Preserving Projections for Multi-Agent Planning [link]Paper  abstract   bibtex   
Collaborative privacy-preserving planning (CPPP) is a multi-agent planning task in which agents need to achieve a common set of goals without revealing certain private information to each other. In many privacy-aware planning algorithms the individual agents reason about a projection of the multi-agent problem onto a single agent classical planning problem. For example, an agent can behave as if it controls the public actions of other agents, ignoring their unknown private preconditions and effects. This projection, however, contains very minimal information, ignoring the dependencies between public actions of other agents. We propose a stronger type of projection, in which the information about the dependencies between the public actions is published to all agents. In addition, we propose a new definition of privacy referring to the number of private facts or objects an agent controls, and show that we do not disclose such information. The benefits of our stronger projection are demonstrated by showing that it produces high level plans very fast, solving more benchmark problems than any other state-of-the-art privacy preserving planners.
@inproceedings {icaps16-17,
    track    = {​Main Track},
    title    = {Stronger Privacy Preserving Projections for Multi-Agent Planning},
    url      = {http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13095},
    author   = {Guy Shani and  Shlomi Maliah and  Roni Stern},
    abstract = {Collaborative privacy-preserving planning (CPPP) is a multi-agent planning task in which agents need to achieve a common set of goals without revealing certain private information to each other. In many privacy-aware planning algorithms the individual agents reason about a projection of the multi-agent problem onto a single agent classical planning problem. 
For example, an agent can behave as if it controls the public actions of other agents, ignoring their unknown private preconditions and effects. This projection, however, contains very minimal information, ignoring the dependencies between public actions of other agents. We propose a stronger type of projection, in which the information about the dependencies between the public actions is published to all agents. In addition, we propose a new definition of privacy referring to the number of private facts or objects an agent controls, and show that we do not disclose such information.
The benefits of our stronger projection are demonstrated by showing that it produces high level plans very fast, solving more benchmark problems than any other state-of-the-art privacy preserving planners.},
    keywords = {Distributed and multi-agent planning}
}

Downloads: 0