The Simplest Multi-key Linearly Homomorphic Signature Scheme

Diego F. Aranha,Elena Pagnin
DOI: https://doi.org/10.1007/978-3-030-30530-7_14
2019-01-01
Abstract:We consider the problem of outsourcing computation on data authenticated by different users. Our aim is to describe and implement the simplest possible solution to provide data integrity in cloud-based scenarios. Concretely, our multi-key linearly homomorphic signature scheme (mklhsdocumentclass[12pt]{minimal}usepackage{amsmath}usepackage{wasysym}usepackage{amsfonts}usepackage{amssymb}usepackage{amsbsy}usepackage{mathrsfs}usepackage{upgreek}setlength{oddsidemargin}{-69pt}egin{document}$${mathsf {mklhs}}$$end{document}) allows users to upload signed data on a server, and at any later point in time any third party can query the server to compute a linear combination of data authenticated by different users and check the correctness of the returned result. Our construction generalizes Boneh et al.’s linearly homomorphic signature scheme (PKC’09 [7]) to the multi-key setting and relies on basic tools of pairing-based cryptography. Compared to existing multi-key homomorphic signature schemes, our mklhsdocumentclass[12pt]{minimal}usepackage{amsmath}usepackage{wasysym}usepackage{amsfonts}usepackage{amssymb}usepackage{amsbsy}usepackage{mathrsfs}usepackage{upgreek}setlength{oddsidemargin}{-69pt}egin{document}$${mathsf {mklhs}}$$end{document} is a conceptually simple and elegant direct construction, which trades-off privacy for efficiency. The simplicity of our approach leads us to a very efficient construction that enjoys significantly shorter signatures and higher performance than previous proposals. Finally, we implement mklhsdocumentclass[12pt]{minimal}usepackage{amsmath}usepackage{wasysym}usepackage{amsfonts}usepackage{amssymb}usepackage{amsbsy}usepackage{mathrsfs}usepackage{upgreek}setlength{oddsidemargin}{-69pt}egin{document}$${mathsf {mklhs}}$$end{document} using two different pairing-friendly curves at the 128-bit security level, a Barreto-Lynn-Scott curve and a Barreto-Naehrig curve. Our benchmarks illustrate interesting performance trade-offs between these parameters, involving the cost of exponentiation and hashing in pairing groups. We provide a discussion on such trade-offs that can be useful to other implementers of pairing-based protocols.
What problem does this paper attempt to address?