William E. Skeith

Private searching on streaming data

In this paper, we consider the problem of private searching on streaming data. We show that in this model we can efficiently implement searching for documents under a secret criteria (such as presence or absence of a hidden combination of hidden keywords) under various cryptographic assumptions. Our results can be viewed in a variety of ways: as a generalization of the notion of a Private Information Retrieval (to the more general queries and to a streaming environment as well as to public-key program obfuscation); as positive results on privacy-preserving datamining; and as a delegation of hidden program computation to other machines.

Private Searching on Streaming Data

In this paper we consider the problem of private searching on streaming data, where we can efficiently implement searching for documents that satisfy a secret criteria (such as the presence or absence of a hidden combination of hidden keywords) under various cryptographic assumptions. Our results can be viewed in a variety of ways: as a generalization of the notion of private information retrieval (to more general queries and to a streaming environment); as positive results on privacy-preserving datamining; and as a delegation of hidden program computation to other machines.

Hard-Core Predicates for a Diffie-Hellman Problem over Finite Fields

A long-standing open problem in cryptography is proving the existence of (deterministic) hard-core predicates for the Diffie-Hellman problem defined over finite fields. In this paper, we make progress on this problem by defining a very natural variation of the Diffie-Hellman problem over \(\mathbb{F}_{p^2}\) and proving the unpredictability of every single bit of one of the coordinates of the secret DH value.

Generalized learning problems and applications to non-commutative cryptography

We propose a generalization of the learning parity with noise (LPN) and learning with errors (LWE) problems to an abstract class of group-theoretic learning problems that we term learning homomorphisms with noise (LHN). This class of problems contains LPN and LWE as special cases, but is much more general. It allows, for example, instantiations based on non-abelian groups, resulting in a new avenue for the application of combinatorial group theory to the development of cryptographic primitives. We then study a particular instantiation using relatively free groups and construct a symmetric cryptosystem based upon it.

Homomorphic Secret Sharing from Paillier Encryption

A recent breakthrough by Boyle et al. [7] demonstrated secure function evaluation protocols for branching programs, where the communication complexity is sublinear in the size of the circuit (indeed just linear in the size of the inputs, and polynomial in the security parameter). Their result is based on the Decisional Diffie-Hellman assumption (DDH), using (variants of) the ElGamal cryptosystem. In this work, we extend their result to show a construction based on the circular security of the Paillier encryption scheme. We also offer a few optimizations to the scheme, including an alternative to the “Las Vegas”-style share conversion protocols of [7, 9] which directly checks the correctness of the computation. This allows us to reduce the number of required repetitions to achieve a desired overall error bound by a constant fraction for typical cases, and for large programs, reduces the total computation cost.

On the Black-box Use of Somewhat Homomorphic Encryption in NonInteractive Two-Party Protocols

In this work, we develop a methodology for determining the communication required to implement various two-party functionalities noninteractively. In the particular setting on which we focus, the protocols are based upon somewhat homomorphic encryption, and furthermore, they treat the homomorphic properties as a black box. In this setting, we develop lower bounds which give a smooth trade-off between the communication complexity and the “expressiveness” of the cryptosystem---the latter being measured in terms of the depth of the arithmetic circuits that can be evaluated on ciphertext. Given the current state of the art in homomorphic encryption, this trade-off may also be viewed as one between communication and computation, since at present, more expressive cryptosystems are markedly less efficient. We then apply this methodology to place lower bounds on a number of cryptographic protocols including private information retrieval writing and private keyword search. Our work provides a useful “litmus test” ...

Hardness of learning problems over Burnside groups of exponent 3

In this work, we investigate the hardness of learning Burnside homomorphisms with noise (

Algebraic Lower Bounds for Computing on Encrypted Data

B_{n} \hbox {-}\mathsf {LHN}