ABOUT

About SmartData

This Symposium brings together a group of international, multi-disciplinary experts, to set in motion research in online privacy and security. Following the model of the 1956 Dartmouth conference on Artificial Intelligence, we expect the Symposium to help determine the future course of “SmartData” research.

Enter SmartData, the subject matter of this Symposium: Internet-based autonomous agents which act as an individual's online surrogate, securely storing personal information, and intelligently disclosing it based upon the context of the data request, and in accordance with the user’s instructions. The goal is to surpass current limited and brittle data protection methods by being able to respond to unforeseen situations, adapt to novel threats, and provide an accurate and nuanced representation of an individual's privacy and data security preferences.

In the 40+ years of its existence, the Internet has evolved from an information distribution service to a powerful instrument for more effective business practices, social interaction, and even political emancipation. Coincident with this evolution has been an immense growth in the demand for personal information — an increasingly valuable commodity — and a decrease in the ability of individuals to maintain informational privacy (control over the use and dissemination of one's personal information). The future of privacy may well depend on the ability of individuals to reclaim personal control of their information and identities online. In failing to protect personal data, the liberating potential of the Internet may be compromised, as various organizations (public, private, and criminal) may use this free flow of information to exert control over, and potentially harm, individuals. We thus face the daunting question: can 21st-century technology be designed to protect personal information and, in the process, support individual liberty and freedom?

The proposed model for SmartData is based upon evolutionary computation and embodied cognition within a dynamical systems framework. As such, it will depend on the evolution of agents, embodied in an interactive and modifiable virtual environment. The advances of the last decade in evolutionary robotics using embodied cognition (in a non-linear dynamical systems model) have provided optimism that 'virtual agents' may be able to securely store personal information and evolve/learn the necessary contexts for the appropriate release of their personally-identifiable information (PII). The unknown at this point is whether the techniques of evolutionary robotics will scale sufficiently to encompass input domains large enough for practical applications.

By empowering (but not burdening) data subjects, and returning to them control of their own data, a positive-sum outcome — in which privacy and the key functionalities of the Internet are achieved — can be realized. In effect, this would return control of personal information to the individual, where it belongs. Further, the autonomous nature and adaptability of the virtual agent allows this control to be maintained without significant burden to the data subject — he or she need only set up the initial conditions for data release contexts (and inform the agent should these conditions change), with the virtual agent then applying these in multiple scenarios. Lastly, these embedded agents will ensure that evolution and adaptability are achieved within the limits and functionalities of the Internet.

The two foundational papers, “SmartData: The Need, the Goal, the Challenge” and “SmartData: Make the data ‘think’ for itself Data protection for the 21st century” by Tomko et al., help introduce SmartData.