Most of the work in the analysis of cryptographic schemes is concentrated in abstract adversarial models that do not capture side-channel attacks. Such attacks exploit various forms of unintended information leakage, which is inherent to almost all physical implementations. Inspired by recent side-channel attacks, especially the “cold boot attacks ” of Halderman et al. (USENIX Security ’08), Akavia, Goldwasser and Vaikuntanathan (TCC ’09) formalized a realistic framework for modeling the security of encryption schemes against a wide class of sidechannel attacks in which adversarially chosen functions of the secret key are leaked. In the setting of public-key encryption, Akavia et al. showed that Regev’s lattice-based scheme (STOC ’05) is resilient to any leakage of

Abstract. We present a somewhat homomorphic encryption scheme that is both very simple to describe and analyze, and whose security (quantumly) reduces to the worst-case hardness of problems on ideal lattices. We then transform it into a fully homomorphic encryption scheme using standard “squashing ” and “bootstrapping ” techniques introduced by Gentry (STOC 2009). One of the obstacles in going from “somewhat ” to full homomorphism is the requirement that the somewhat homomorphic scheme be circular secure, namely, the scheme can be used to securely encrypt its own secret key. For all known somewhat homomorphic encryption schemes, this requirement was not known to be achievable under any cryptographic assumption, and had to be explicitly assumed. We take a step forward towards removing this additional assumption by proving that our scheme is in fact secure when encrypting polynomial functions of the secret key. Our scheme is based on the ring learning with errors (RLWE) assumption that was recently introduced by Lyubashevsky, Peikert and Regev (Eurocrypt 2010). The RLWE assumption is reducible to worst-case problems on ideal lattices, and allows us to completely abstract out the lattice interpretation, resulting in an extremely simple scheme. For example, our secret key is s, and our public key is (a, b = as + 2e), where s, a, e are all degree (n − 1) integer polynomials whose coefficients are independently drawn from easy to sample distributions. 1

Key-dependent message (KDM) secure encryption schemes provide secrecy even when the attacker sees encryptions of messages related to the secret-key sk. Namely, the scheme should remain secure even when messages of the form f(sk) are encrypted, where f is taken from some function class F. A KDM amplification procedure takes an encryption scheme which satisfies F-KDM security and boost it into a G-KDM secure scheme, where the function class G should be richer than F. It was recently shown by Brakerski et al. (TCC 2011) and Barak et al. (EUROCRYPT 2010), that a strong form of amplification is possible, provided that the underlying encryption scheme satisfies some special additional properties. In this work, we prove the first generic KDM amplification theorem which relies solely on the KDM security of the underlying scheme without making any other assumptions. Specifically, we show that an elementary form of KDM security against functions in which each output bit either copies or flips a single bit of the key (aka projections) can be amplified into KDM security with respect to any function family that can be computed in arbitrary fixed polynomial-time. Furthermore, our amplification theorem and its proof are insensitive to the exact setting of KDM security, and they hold in the presence of multiple-keys and in the symmetric-key/public-key and the CPA/CCA cases. As a result, we can amplify the security of all known KDM constructions, including ones that could not be amplified before. Finally, we study the minimal conditions under which full-KDM security (with respect to all functions) can be achieved. We show that under strong notion of KDM security, the existence of cyclic-secure fully-homomorphic encryption is not only sufficient for full-KDM security, as shown by Barak et al., but also necessary. On the other hand, we observe that for standard KDM security, this condition can be relaxed by adopting Gentry’s bootstrapping technique (STOC 2009) to the KDM setting.

In this work, we show that strong leakage resilience for cryptosystems with advanced functionalities can be obtained quite naturally within the methodology of dual system encryption, recently introduced by Waters. We demonstrate this concretely by providing fully secure IBE, HIBE, and ABE systems which are resilient to bounded leakage from each of many secret keys per user, as well as many master keys. This can be realized as resilience against continual leakage if we assume keys are periodically updated and no (or logarithmic) leakage is allowed during the update process. Our systems are obtained by applying a simple modification to previous dual system encryption constructions: essentially this provides a generic tool for making dual system encryption schemes leakage-resilient. 1

A signature scheme is fully leakage resilient (Katz and Vaikuntanathan, ASIACRYPT ’09) if it is existentially unforgeable under an adaptive chosen-message attack even in a setting where an adversary may obtain bounded (yet arbitrary) leakage information on all intermediate values that are used throughout the lifetime of the system. This is a strong and meaningful notion of security that captures a wide range of side-channel attacks. One of the main challenges in constructing fully leakage-resilient signature schemes is dealing with leakage that may depend on the random bits used by the signing algorithm, and constructions of such schemes are known only in the random-oracle model. Moreover, even in the random-oracle model, known schemes are only resilient to leakage of less than half the length of their signing key. In this paper we construct the first fully leakage-resilient signature schemes without random oracles. We present a scheme that is resilient to any leakage of length (1 − o(1))L bits, where L is the length of the signing key. Our approach relies on generic cryptographic primitives, and at the same time admits rather efficient instantiations based on specific number-theoretic

We exhibit a hash-based storage auditing scheme which is provably secure in the random-oracle model (ROM), but easily broken when one instead uses typical indifferentiable hash constructions. This contradicts the widely accepted belief that the indifferentiability composition theorem applies to any cryptosystem. We characterize the uncovered limitation of the indifferentiability framework by showing that the formalizations used thus far implicitly exclude security notions captured by experiments that have multiple, disjoint adversarial stages. Examples include deterministic public-key encryption (PKE), password-based cryptography, hash function nonmalleability, key-dependent message security, and more. We formalize a stronger notion, reset indifferentiability, that enables an indifferentiabilitystyle composition theorem covering such multi-stage security notions, but then show that practical hash constructions cannot be reset indifferentiable. We discuss how these limitations also affect the universal composability framework. We finish by showing the chosen-distribution attack security (which requires a multi-stage game) of some important public-key encryption schemes built using a hash construction paradigm introduced by Dodis, Ristenpart, and Shrimpton. 1

Deterministic public-key encryption, introduced by Bellare, Boldyreva, and O’Neill (CRYPTO ’07), provides an alternative to randomized public-key encryption in various scenarios where the latter exhibits inherent drawbacks. A deterministic encryption algorithm, however, cannot satisfy any meaningful notion of security when the plaintext is distributed over a small set. Bellare et al. addressed this difficulty by requiring semantic security to hold only when the plaintext has high min-entropy from the adversary’s point of view. In many applications, however, an adversary may obtain auxiliary information that is related to the plaintext. Specifically, when deterministic encryption is used as a building block of a larger system, it is rather likely that plaintexts do not have high min-entropy from the adversary’s point of view. In such cases, the framework of Bellare et al. might fall short from providing robust security guarantees. We formalize a framework for studying the security of deterministic public-key encryption schemes with respect to auxiliary inputs. Given the trivial requirement that the plaintext should not be efficiently recoverable from the auxiliary input, we focus on hard-to-invert auxiliary inputs.

In the continual memory leakage model, security against attackers who can repeatedly obtain leakage is achieved by periodically updating the secret key. This is an appealing model which captures a wide class of side-channel attacks, but all previous constructions in this model provide only a very minimal amount of leakage tolerance during secret key updates. Since key updates may happen frequently, improving security guarantees against attackers who obtain leakage during these updates is an important problem. In this work, we present the first cryptographic primitives which are secure against a super-logarithmic amount of leakage during secret key updates. We present signature and public key encryption schemes in the standard model which can tolerate a constant fraction of the secret key to be leaked between updates as well as a constant fraction of the secret key and update randomness to be leaked during updates. Our signature scheme also allows us to leak a constant fraction of the entire secret state during signing. Before this work, it was unknown how to tolerate super-logarithmic leakage during updates even in the random oracle model. We rely on subgroup decision assumptions in composite order bilinear groups. 1

Abstract. A cryptographic primitive is leakage-resilient, if it remains secure even if an adversary can learn a bounded amount of arbitrary information about the computation with every invocation. As a consequence, the physical implementation of a leakage-resilient primitive is secure against every side-channel as long as the amount of information leaked per invocation is bounded. In this paper we prove positive and negative results about the feasibility of constructing leakage-resilient pseudorandom functions and permutations (i.e. block-ciphers). Our results are three fold: 1. We construct (from any standard PRF) a PRF which satisfies a relaxed notion of leakageresilience where (1) the leakage function is fixed (and not adaptively chosen with each query.) and (2) the computation is split into several steps which leak individually (a “step ” will be the invocation of the underlying PRF.) 2. We prove that a Feistel network with a super-logarithmic number of rounds, each instantiated with a leakage-resilient PRF, is a leakage resilient PRP. This reduction also holds for the non-adaptive notion just discussed, we thus get a block-cipher which is leakage-resilient (against non-adaptive leakage). 3. We propose generic side-channel attacks against Feistel networks. The attacks are generic in the sense that they work for any round functions (e.g. uniformly random functions) and only require some simple leakage from the inputs to the round functions. For example we show how to invert an r round Feistel network over 2n bits making 4 · (n + 1) r−2 forward queries, if with each query we are also given as leakage the Hamming weight of the inputs to the r round functions. This complements the result from the previous item showing that a super-constant number of rounds is necessary. 1

Yao’s garbled-circuit approach enables constant-round secure two-party computation for any boolean circuit. In Yao’s original construction, each gate in the circuit requires the parties to perform a constant number of encryptions/decryptions, and to send/receive a constant number of ciphertexts. Kolesnikov and Schneider (ICALP 2008) proposed an improvement that allows XOR gates in the circuit to be evaluated “for free”, i.e., incurring no cryptographic operations and zero communication. Their “free-XOR ” technique has proven very popular, and has been shown to improve performance of garbled-circuit protocols by up to a factor of 4. Kolesnikov and Schneider proved security of their approach in the random oracle model, and claimed that (an unspecified variant of) correlation robustness would suffice; this claim has been repeated in subsequent work, and similar ideas have since been used (with the same claim about correlation robustness) in other contexts. We show that, in fact, the free-XOR technique cannot be proven secure based on correlation robustness alone: somewhat surprisingly, some form of circular security is also required. We propose an appropriate notion of security for hash functions capturing the necessary requirements, and prove security of the free-XOR approach when instantiated with any hash function satisfying our definition. Our results do not impact the security of the free-XOR technique in practice, or imply an error in the free-XOR work, but instead pin down the assumptions needed to prove security.