Cryptographic protocols are commonly designed and their security proven under the assumption that the protocol parties have access to perfect (uniform) randomness. Physical randomness sources deployed in practical implementations of these protocols often fall short in meeting this assumption, but instead provide only a steady stream of bits with certain high entropy. Trying to ground cryptographic protocols on such imperfect, weaker sources of randomness has thus far mostly given rise to a multitude of impossibility results, including the impossibility to construct provably secure encryption, commitments, secret sharing, and zero-knowledge proofs based solely on a weak source. More generally, indistinguishability-based properties break down for such weak sources. In this paper, we show that the loss of security induced by using a weak source can be meaningfully quantified if the source is bounded, e.g., for the well-studied Santha-Vazirani (SV) sources. The quantification relies on a novel relaxation of indistinguishability by a quantitative parameter. We call the resulting notion dierential indistinguishability in order to reflect its structural similarity to dierential privacy. More concretely, we prove that indistinguishability with uniform randomness implies dierential indistinguishability with weak randomness. We show that if the amount of weak randomness is limited (e.g., by using it only to seed a PRG), all cryptographic primitives and protocols still achieve dierential indistinguishability.
Cryptographic protocols are commonly designed and their security proven under the assumption that the protocol parties have access to perfect (uniform) randomness. Physical randomness sources deployed in practical implementations of these protocols often fall short in meeting this assumption, but instead provide only a steady stream of bits with certain high entropy. Trying to ground cryptographic protocols on such imperfect, weaker sources of randomness has thus far mostly given rise to a multitude of impossibility results, including the impossibility to construct provably secure encryption, commitments, secret sharing, and zero-knowledge proofs based solely on a weak source. More generally, indistinguishability-based properties break down for such weak sources. In this paper, we show that the loss of security induced by using a weak source can be meaningfully quantified if the source is bounded, e.g., for the well-studied Santha-Vazirani (SV) sources. The quantification relies on a novel relaxation of indistinguishability by a quantitative parameter. We call the resulting notion dierential indistinguishability in order to reflect its structural similarity to dierential privacy. More concretely, we prove that indistinguishability with uniform randomness implies dierential indistinguishability with weak randomness. We show that if the amount of weak randomness is limited (e.g., by using it only to seed a PRG), all cryptographic primitives and protocols still achieve dierential indistinguishability.
We investigate the feasibility of a variety of cryptographic tasks with imperfect randomness. The kind of imperfect randomness we consider are entropy sources, such as those considered by Santha and Vazirani, Chor and Goldreich, and Zuckerman. We show the following: (1) certain cryptographic tasks like bit commitment, encryption, secret sharing, zero-knowledge, non-interactive zero-knowledge, and secure two-party computation for any non-trivial junction are impossible to realize if parties have access to entropy sources with slightly less-than-perfect entropy, i.e., sources with imperfect randomness. These results are unconditional and do not rely on any un-proven assumption. (2) On the other hand, based on stronger variants of standard assumptions, secure signature schemes are possible with imperfect entropy sources. As another positive result, we show (without any unproven assumption) that interactive proofs can be made sound with respect to imperfect entropy sources.
Randomness plays a central role in computer science, in particular cryptography. Almost all cryptographic primitives depend crucially on randomness because randomness and unpredictability in secret keys provides the means for security. Usually one assumes that perfect randomness, a sequence of independently and uniformly distributed bits, are accessible to algorithms. This is a strong assumption. Physical sources of randomness are neither uniformly random, nor produce necessarily independent bits. Therefore, the aim of this thesis is to start from a realistic model of randomness, investigate notions of secrecy and their randomness requirements and finally find practical methods for generation of randomness that matches the requirements of cryptographic primitives. We consider a model of random source where the source output follows one distribution from a set of possible distributions, each with the property that the maximum probability of symbols is bounded and can not be arbitrarily close to 1. This model does not assume independence or uniformity of the output symbols, and is considered to be a realistic model of randomness. From this point, the thesis can be divided into two main parts: In the first part, considering various notions of information theoretic secrecy, a fundamental problem is to find the properties of randomness needed to achieve security in these notions. Traditional cryptographic protocols simply assume perfect randomness and build on this assumption. We explore the results that show secrecy can not be based on imperfect randomness that is not uniform or independent. Thus a line of work attempts to relax notions of secrecy in such a way that they can be constructed with non-perfect sources, and possibly require smaller key sizes. Yet they should match real life applications. An important work in this context is entropic security where the key size could be smaller than the message depending on message distribution. Inspired by this, we propose two relaxed notions of secrecy that are motivated by practical applications. In the first notion, motivated by an application in biometrics authentication, we propose guessing secrecy where the probability of guessing the message for a computational unbounded adversary with the best strategy remains the same when a ciphertext is given or when it is not. We compare the randomness requirements of guessing secrecy with stronger notions and show that in some cases such as key length, the requirements are the same. For key distribution however, we found a family of distributions that provide guessing secrecy but not perfect secrecy. In the second notion, we investigate randomness requirements of multiple message encryption. Considering a natural extension of secrecy definition to multiple messages, we show that independent keys are needed to encrypt each message. We then propose a relaxed notion in which the security of last message is more important than past messages, although the leakage of past messages is bounded using entropic security. By this assumption, we achieve smaller key length compared to indistinguishability, and comparable to key length for entropic security. This notion has applications such as location privacy. In the second part of the thesis, since secrecy crucially depends on perfect randomness, we investigate how perfect randomness can be practically generated, specifically from human game-play. Unlike many random number generators that assume independent or uniform random bits in the random source, we base our work on the realistic model of randomness. Our main observation is that human game-play has an element of randomness coming from the errors in their game-play which is the main entertaining factor of the game. We also observed that this game-play can distinguish among a group of people if the right features are collected from the game-play. We incrementally changed our game design until the distinguishability among a small population is maximized, and then run the experiments required to show the viability of this approach over a larger population. This approach can also provide a hard to delegate authentication property where a human could not emulate the behavior of another human even given statistical information about their game-play.
We revisit the impossibility of a variety of cryptographic tasks including privacy and differential privacy with imperfect randomness. For traditional notions of privacy, such as security of encryption, commitment or secret sharing schemes, dramatic impossibility results are known [MP90,DOPS04] for several concrete sources R, including a (seemingly) very “nice and friendly” Santha-Vazirani (SV) source. Moreover, Bosley and Dodis [BD07] gave strong evidence that many traditional privacy tasks (e.g., encryption) inherently require an “extractable” source of randomness. The common interpretation of these negative results is that privacy is impossible even with “very structured” imperfect sources. Somewhat surprisingly, Dodis et al. [DLMV12] put a slight dent in this belief, by showing that non-trivial differential privacy is possible with the SV sources. This suggested a qualitative gap between traditional and differential privacy, and left open the question if differential privacy is possible with more realistic (i.e., less structured) sources than the SV sources. Motivated by solving this question, we introduce a new, modular framework for showing strong impossibility results for (either traditional or differential) privacy under a general imperfect source R. In particular, we introduce natural and easy-to-understand concepts of expressiveness and separability of a given imperfect source R, and show the following results: – Low levels of expressiveness generically imply strong impossibility results for both traditional and differential privacy. – Separability implies expressiveness; NON-separability is equivalent to “weak bit extraction”. – While the separability of some concrete (e.g., SV) sources R was implicitly known, we show new separability results for several important sources, including general “block sources”. As direct corollaries of these general results, we get the following corollaries: – Existing, but quantitatively improved, impossibility results for traditional privacy, but under a wider variety of sources R. – First impossibility results for differential privacy. Although, unsurprisingly, these results (barely) miss the highly structured SV sources, they come back extremely quickly once the source becomes slightly more realistic (e.g., a realistic “block source”). – Any imperfect source allowing (either traditional or differential) privacy admits a certain type of deterministic bit extraction. (This result is incomparable to the result of [BD07].) Overall, we believe our results provide an intuitive, modular and unified picture elucidating the (im)possibility of privacy with general imperfect sources.
There has been much recent progress in constructing cryptosystems that maintain their security without requiring uniform randomness and perfect secrecy. These schemes are motivated by a diverse set of problems such as providing resilience to side-channel leakage, using weak physical sources of randomness as secret keys, and allowing deterministic encryption for high-entropy messages. The study of these problems has significantly deepened our understanding of how randomness is used in cryptographic constructions and proofs. Nevertheless, despite this progress, some basic and seemingly achievable security properties have eluded our reach. For example, we are unable to prove the security of basic tools for manipulating weak/leaky random sources, such as as pseudo-entropy generators and seed-dependent computational condensers. We also do not know how to prove leakage-resilient security of any cryptosystem with a uniquely determined secret key. In the context of deterministic encryption we do not have a standardmodel constructions achieving the strongest notion of security proposed by Bellare, Boldyreva and O’Neill (CRYPTO ’07), that would allow us to encrypt arbitrarily correlated messages of sufficiently large individual entropy. In this work, we provide broad black-box separation results, showing that the security of such primitives cannot be proven under virtually any standard cryptographic hardness assumption via a reduction that treats the adversary as a black box. We do so by formalizing the intuition that “the only way that a reduction can simulate the correctly distributed view for an attacker is to know all the secrets, in which case it does not learn anything useful from the attack”. Such claims are often misleading and clever way of getting around them allow us to achieve a wealth of positive results with imperfect/leaky randomness. However, in this work we show that this intuition can be formalized and that it indeed presents a real barrier for the examples given above. ∗IBM Research. wichs@cs.nyu.edu
Abstract. Most cryptographic primitives such as encryption, authentication or secret sharing require randomness. Usually one assumes that perfect randomness is available, but those primitives might also be realized under weaker assumptions. In this work we continue the study of building secure cryptographic primitives from imperfect random sources initiated by Dodis and Spencer (FOCS’02). Their main result shows that there exists a (high-entropy) source of randomness allowing for perfect encryption of a bit, and yet from which one cannot extract even a single weakly random bit, separating encryption from extraction. Our main result separates encryption from 2-out-2 secret sharing (both in the information-theoretic and in the computational settings): any source which can be used to achieve one-bit encryption also can be used for 2-out-2 secret sharing of one bit, but the converse is false, even for highentropy sources. Therefore, possibility of extraction strictly implies encryption, which in turn strictly implies 2-out-2 secret sharing. 1
Cryptographic protocols, though used in a wide range of everyday tasks, are typically built for certain unrealistic ideal conditions in which the user is assumed to work. For example, one such condition that is implicitly assumed is that the user has perfect randomness that she can use as her cryptographic secret key. Another commonly made assumption is that the user runs her cryptographic tasks (such as encryption) on a device that is perfectly secure; that is, its internals are not tampered with or even observed by an adversary.
In this thesis, we focus on designing cryptographic techniques that are secure under less-than-ideal conditions. In order to remove the assumption that users share perfectly random secret keys, we study the problem of privacy amplification: key agreement between two parties who both know a weak secret w, such as a password. The goal of the protocol is to convert this non-uniform secret w into a uniformly distributed string R that is fully secret from the adversary. R may then be used as a key for running symmetric cryptographic protocols (such as encryption or authentication). Because we make no computational assumptions, the entropy in R can come only from w. Thus such a protocol must minimize the entropy loss during its execution, so that R is as long as possible. In this thesis, we present the first protocol for information-theoretic key agreement that has entropy loss linear in the security parameter. The result is optimal up to constant factors. We also extend our results to the case where the users share correlated secrets w and w' . This setting occurs, for example, when the secrets are biometric readings taken at different points of time and is more generally known as information-reconciliation.