Ph.D. Proposal: TrueBahn - A Trustworthy Data Management Framework for Pervasive Environments

Emerging low power wireless technologies, processors, and abundant storage are enabling devices with very small form factors to host complex services. Personal mobile devices are serving multi-functional roles, from location tracking (GPS) to micro-transactions (smart-cards). Continuing proliferation of such devices will be manifested in resource-rich environments, where information is available locally and directly -- via wireless means, instead of (or in addition to) delivery over infrastructure-based networks. Such pervasive environments are in a constant state of flux. Finding useful services and data is serendipitous, since continuous connectivity to reliable information sources cannot be guaranteed. Ability to form ad hoc networks depends on the participants' willingness in collaborating to relay information on behalf of each other. Data reliability is also an important issue, since prior trust relationships may not always exist. Moreover, finding reliable data on-demand in a serendipitous environment is another challenge. Past research has provided insights into how a profile-driven agent can use notions of trust and reputations for query processing. However, several challenges remain, viz., how do we find reliable information, use active collaborations, leverage abundant storage -- to collate and coordinate data search, and improve latency of simplistic discovery mechanisms. Data sources can be varied and distributed; currently no basic primitives exist to retrieve and evaluate the reliability of data and trustworthiness of the source. It is important to quickly differentiate between trusted and unreliable sources with minimal effort. We propose to achieve this through mutual collaboration--"pack formation." Privileges accorded to a pack member include coordinated queries and trust recommendations, which lead to improved query response times and success rates. Incentives must be provided to offset the costs involved in collaboration, viz., storage and message transmissions. We propose to provide basic primitives to enable on-demand goal-oriented collaborations for sharing data and resources in ad hoc environments, with practical considerations for privacy, trust, and reliability of the resources shared. We propose to design algorithms for achieving ``pack formation'' and metrics for evaluating trust in such associations. Further, we propose to create a data(resource) management framework that employs these primitives and algorithms, and that will facilitate fast and inexpensive data hoarding, with strong assurance of reliability and freshness of data, based on trust-assisted data retrievals and updates. An ensemble of devices collaborating in a resource-rich environment will provide a trustworthy data routing substrate that would support several interesting applications including location-tracking, surveillance, and coordinated information retrieval.