Abstract. In this paper we describe the design and implementation of cloudbased assured information sharing systems. In particular, we will describe our current implementation of a centralized cloud-based assured information sharing system and the design of a decentralized hybrid cloud-based assured information sharing system of the future. Our goal is for coalition organizations to share information stored in multiple clouds and enforce appropriate policies.
IntroductionThe advent of cloud computing and the continuing movement toward software as a service (SaaS) paradigms has posed an increasing need for assured information sharing (AIS) as a service in the cloud. The urgency of this need has been voiced as recently as April 2011 by NSA CIO Lonny Anderson in describing the agency's focus on a "cloud-centric" approach to information sharing with other agencies [1]. Likewise, the DoD has been embracing cloud computing paradigms to more efficiently, economically, flexibly, and scalably meet its vision of "delivering the power of information to ensure mission success through an agile enterprise with freedom of maneuverability across the information environment" [2][3][4][5]. Both agencies therefore have a tremendous need for effective AIS technologies and tools for cloud environments. Although a number of AIS tools have been developed over the past five years for policy-based information sharing [5][6][7][8], to our knowledge none of these tools operate in the cloud and hence do not provide the scalability needed to support large numbers of users utilizing massive amounts of data. Recent prototype systems for supporting cloud-based AIS have applied cloud-centric engines that query large amounts of data in relational databases via non-cloud policy engines that enforce policies expressed in XACML [9][10]. While this is a significant improvement over prior efforts (and has given us insights into implementing cloud-based solutions), it nevertheless has at least three significant limitations. First, XACML-based policy specifications are not expressive enough to support many of the complex policies needed for AIS missions like those of the NSA and DoD. Second, to meet the scalability and efficiency requirements of mission-critical tasks, the policy engine needs to operate in the cloud rather than externally. Third, secure query processing based on relational technology has limitations in representing and processing unstructured data needed for command and control applications.