Group signatures are signatures providing signer anonymity where signers can produce signatures on behalf of the group that they belong to. Although such anonymity is quite attractive considering privacy issues, it is not trivial to check whether a signer has been revoked or not. Thus, how to revoke the rights of signers is one of the major topics in the research on group signatures. In particular, scalability, where the signing and verification costs and the signature size are constant in terms of the number of signers N, and other costs regarding signers are at most logarithmic in N, is quite important. In this paper, we propose a revocable group signature scheme which is currently more efficient compared to previous all scalable schemes. Moreover, our revocable group signature scheme is secure under simple assumptions (in the random oracle model), whereas all scalable schemes are secure under q-type assumptions. We implemented our scheme by employing a Barreto-Lynn-Scott curve of embedding degree 12 over a 455-bit prime field (BLS-12-455), and a Barreto-Naehrig curve of embedding degree 12 over a 382-bit prime field (BN-12-382), respectively, by using the RELIC library. We showed that the online running times of our signing algorithm were approximately 14msec (BLS-12-455) and 11msec (BN-12-382), and those of our verification algorithm were approximately 20msec (BLS-12-455) and 16msec (BN-12-382), respectively. Finally, we showed that our scheme (with a slight extension) is applied to an identity management system proposed by Isshiki et al.

The security of pairing-based cryptosystems is determined by the difficulty of solving the discrete logarithm problem (DLP) over certain types of finite fields. One of the most efficient algorithms for computing a pairing is the ηT pairing over supersingular curves on finite fields of characteristic 3. Indeed many high-speed implementations of this pairing have been reported, and it is an attractive candidate for practical deployment of pairing-based cryptosystems. Since the embedding degree of the ηT pairing is 6, we deal with the difficulty of solving a DLP over the finite field GF(36n), where the function field sieve (FFS) is known as the asymptotically fastest algorithm of solving it. Moreover, several efficient algorithms are employed for implementation of the FFS, such as the large prime variation. In this paper, we estimate the time complexity of solving the DLP for the extension degrees n=97, 163, 193, 239, 313, 353, and 509, when we use the improved FFS. To accomplish our aim, we present several new computable estimation formulas to compute the explicit number of special polynomials used in the improved FFS. Our estimation contributes to the evaluation for the key length of pairing-based cryptosystems using the ηT pairing.

Pairings on elliptic curves over finite fields are crucial for constructing various cryptographic schemes. The ηT pairing on supersingular curves over GF(3n) is particularly popular since it is efficiently implementable. Taking into account the Menezes-Okamoto-Vanstone attack, the discrete logarithm problem (DLP) in GF(36n) becomes a concern for the security of cryptosystems using ηT pairings in this case. In 2006, Joux and Lercier proposed a new variant of the function field sieve in the medium prime case, named JL06-FFS. We have, however, not yet found any practical implementations on JL06-FFS over GF(36n). Therefore, we first fulfill such an implementation and we successfully set a new record for solving the DLP in GF(36n), the DLP in GF(36·71) of 676-bit size. In addition, we also compare JL06-FFS and an earlier version, named JL02-FFS, with practical experiments. Our results confirm that the former is several times faster than the latter under certain conditions.

Logistic regression is a powerful machine learning tool to classify data. When dealing with sensitive or private data, cares are necessary. In this paper, we propose a secure system for privacy-protecting both the training and predicting data in logistic regression via homomorphic encryption. Perhaps surprisingly, despite the non-polynomial tasks of training and predicting in logistic regression, we show that only additively homomorphic encryption is needed to build our system. Indeed, we instantiate our system with Paillier, LWE-based, and ring-LWE-based encryption schemes, highlighting the merits and demerits of each instantiation. Besides examining the costs of computation and communication, we carefully test our system over real datasets to demonstrate its utility.

Due to the legal reform on the protection of personal information in US/Japan and the enforcement of the General Data Protection Regulation (GDPR) in Europe, service providers are obliged to more securely manage the sensitive data stored in their server. In order to protect this kind of data, they generally employ a cryptographic encryption scheme and secure key management schemes such as a Hardware Security Module (HSM) and Trusted Platform Module (TPM). In this paper, we take a different approach based on the space-hard cipher. The space-hard cipher has an interesting property called the space hardness. Space hardness guarantees sufficient security against the adversary who gains a part of key data, e.g., 1/4 of key data. Combined with a simple network monitoring technique, we develop a practical leakage resilient scheme Virtual Vault, which is secure against the snapshot adversary who has full access to the memory in the server for a short period. Importantly, Virtual Vault is deployable by only a low-price device for network monitoring, e.g. L2 switch, and software of space-hard ciphers and packet analyzer, while typical solutions require a dedicated hardware for secure key managements such as HSM and TPM. Thus, Virtual Vault is easily added on the existing servers which do not have such dedicated hardware.

We present the concept of key-rotatable and security-updatable homomorphic encryption (KR-SU-HE) scheme, which is defined as a class of public-key homomorphic encryption in which the keys and the security of any ciphertext can be rotated and updated while still keeping the underlying plaintext intact and unrevealed. After formalising the syntax and security notions for KR-SU-HE schemes, we build a concrete scheme based on the Learning With Errors assumption. We then perform several careful implementations and optimizations to show that our proposed scheme is efficiently practical.

We build a privacy-preserving system of linear regression protecting both input data secrecy and output privacy. Our system achieves those goals simultaneously via a novel combination of homomorphic encryption and differential privacy dedicated to linear regression and its variants (ridge, LASSO). Our system is proved scalable over cloud servers, and its efficiency is extensively checked by careful experiments.

Collecting and analyzing personal data is important in modern information applications. Though the privacy of data providers should be protected, the need to track certain data providers often arises, such as tracing specific patients or adversarial users. Thus, tracking only specific persons without revealing normal users' identities is quite important for operating information systems using personal data. It is difficult to know in advance the rules for specifying the necessity of tracking since the rules are derived by the analysis of collected data. Thus, it would be useful to provide a general way that can employ any data analysis method regardless of the type of data and the nature of the rules. In this paper, we propose a privacy-preserving data analysis construction that allows an authority to detect specific users while other honest users are kept anonymous. By using the cryptographic techniques of group signatures with message-dependent opening (GS-MDO) and public key encryption with non-interactive opening (PKENO), we provide a correspondence table that links a user and data in a secure way, and we can employ any anonymization technique and data analysis method. It is particularly worth noting that no “big brother” exists, meaning that no single entity can identify users who do not provide anomaly data, while bad behaviors are always traceable. We show the result of implementing our construction. Briefly, the overhead of our construction is on the order of 10 ms for a single thread. We also confirm the efficiency of our construction by using a real-world dataset.