Visible to the public Attribute-based Key Exchange with General Policies

TitleAttribute-based Key Exchange with General Policies
Publication TypeConference Paper
Year of Publication2016
AuthorsKolesnikov, Vladimir, Krawczyk, Hugo, Lindell, Yehuda, Malozemoff, Alex, Rabin, Tal
Conference NameProceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security
PublisherACM
Conference LocationNew York, NY, USA
ISBN Number978-1-4503-4139-4
Keywordsauthentication, Key exchange, pubcrawl, Resiliency, Scalability, signature based defense, TLS
Abstract

Attribute-based methods provide authorization to parties based on whether their set of attributes (e.g., age, organization, etc.) fulfills a policy. In attribute-based encryption (ABE), authorized parties can decrypt, and in attribute-based credentials (ABCs), authorized parties can authenticate themselves. In this paper, we combine elements of ABE and ABCs together with garbled circuits to construct attribute-based key exchange (ABKE). Our focus is on an interactive solution involving a client that holds a certificate (issued by an authority) vouching for that client's attributes and a server that holds a policy computable on such a set of attributes. The goal is for the server to establish a shared key with the client but only if the client's certified attributes satisfy the policy. Our solution enjoys strong privacy guarantees for both the client and the server, including attribute privacy and unlinkability of client sessions. Our main contribution is a construction of ABKE for arbitrary circuits with high (concrete) efficiency. Specifically, we support general policies expressible as boolean circuits computed on a set of attributes. Even for policies containing hundreds of thousands of gates the performance cost is dominated by two pairing computations per policy input. Put another way, for a similar cost to prior ABE/ABC solutions, which can only support small formulas efficiently, we can support vastly richer policies. We implemented our solution and report on its performance. For policies with 100,000 gates and 200 inputs over a realistic network, the server and client spend 957 ms and 176 ms on computation, respectively. When using offline preprocessing and batch signature verification, this drops to only 243 ms and 97 ms.

URLhttp://doi.acm.org/10.1145/2976749.2978359
DOI10.1145/2976749.2978359
Citation Keykolesnikov_attribute-based_2016