Abstract-TrustedDB is an outsourced database prototype that allows clients to execute SQL queries with privacy and under regulatory compliance constraints without having to trust the service provider. TrustedDB achieves this by leveraging server-hosted tamper-proof trusted hardware in critical query processing stages.TrustedDB does not limit the query expressiveness of supported queries. And, despite the cost overhead and performance limitations of trusted hardware, the costs per query are orders of magnitude lower than any (existing or) potential future software-only mechanisms. TrustedDB is built and runs on actual hardware, and its performance and costs are evaluated here.
I. OVERVIEWOutsourcing has finally arrived, due in no small part to the availability of cheap high speed networks, storage and CPUs. Clients can now minimize their management overheads and virtually eliminate infrastructure costs Virtually all major "cloud" providers today offer a database service of some kind as part of their overall solution. Numerous startups also feature more targeted data management and/or database platforms.Yet, significant challenges lie in the path of large-scale adoption. Such services often require their customers to inherently trust the provider with full access to the outsourced datasets. But numerous instances of illicit insider behavior or data leaks have left clients reluctant to place sensitive data under the control of a remote, third-party provider, without practical assurances of privacy and confidentiality -especially in business, healthcare and government frameworks. And today's privacy guarantees of such services are at best declarative and subject customers to unreasonable fine-print clauses -e.g., allowing the server operator (or malicious attackers gaining access to its systems) to use customer behavior and content for commercial, profiling, or governmental surveillance purposes [5,6].Existing research addresses several such outsourcing security aspects, including access privacy, searches on encrypted data, range queries, and aggregate queries. To achieve privacy, in most of these efforts data is encrypted before outsourcing. Once encrypted however, inherent limitations in the types of primitive operations that can be performed on encrypted data lead to fundamental expressiveness and practicality constraints.Recent theoretical cryptography results provide hope by proving the existence of universal homomorphisms, i.e., encryption mechanisms that allow computation of arbitrary functions without decrypting the inputs [12]. Unfortunately actual instances of such mechanisms seem to be decades away from being practical [7].Ideas have also been proposed to leverage tamper-proof hardware to privately process data server-side, ranging from smartcard deployment [9] in healthcare, to more general database operations [3,8,10].Yet, common wisdom so far has been that trusted hardware is generally impractical due to its performance limitations and higher acquisition costs. As a result, with very few exceptions [9]...