Takuya Hayashi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Takuya Hayashi has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Lihua Wang
National Institute of Information and Communications Technology (NICT)
11 shared publications Yoshinori Aono
National Institute of Information and Communications Technology, Tokyo, Japan
8 shared publicationsLe Trieu Phong
National Institute of Information and Communications Technology, Japan
6 shared publicationsTsuyoshi Takagi
Kyushu University
6 shared publicationsNaoyuki Shinohara
National Institute of Information and Communications Technology
5 shared publications- Keita Emura
Kanazawa University
4 shared publications
Latest Publications
- 13 March 2025
- Published by Springer Nature
- p. 84-104
Multi-key Homomorphic Encryption with Threshold Re-encryption
- 13 March 2025
- Published by Springer Nature
- p. 84-104
Multi-key Homomorphic Encryption with Threshold Re-encryption
- Open Access
- 19 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 76-86
Faster Homomorphic Evaluation of Arbitrary Bivariate Integer Functions via Homomorphic Linear Transformation
Fully homomorphic encryption (FHE) can perform computations on encrypted data, allowing us to analyze sensitive data while maintaining security. Several popular FHE schemes, such as BGV and BFV, are suitable for arithmetic circuits. However, there is still much room for improving the efficiency of the evaluation of non-linear functions in those schemes. We propose an efficient method for the evaluation of arbitrary bivariate functions in homomorphic encryption with integers as plaintext. Prior work on arbitrary bivariate function evaluation for relatively large plaintext integers has been proposed by Maeda et al. (WAHC'22). In the method of Maeda et al., the polynomial evaluation is the most time-consuming part in terms of the execution time. In this work, we construct a method for arbitrary bivariate functions based on homomorphic linear transformations, eliminating the need for polynomial evaluation to enhance efficiency. We also provide a time complexity analysis showing that our method achieves lower time computational complexity. In addition, we implement the previous method of Maeda et al. and our proposed method using OpenFHE with the input domain {0, 1,..., 2^15 - 1} and demonstrate that our method achieves a 1.38 times speed up for the division function compared to the previous method.
- 19 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 76-86
Open AccessFaster Homomorphic Evaluation of Arbitrary Bivariate Integer Functions via Homomorphic Linear Transformation
Fully homomorphic encryption (FHE) can perform computations on encrypted data, allowing us to analyze sensitive data while maintaining security. Several popular FHE schemes, such as BGV and BFV, are suitable for arithmetic circuits. However, there is still much room for improving the efficiency of the evaluation of non-linear functions in those schemes. We propose an efficient method for the evaluation of arbitrary bivariate functions in homomorphic encryption with integers as plaintext. Prior work on arbitrary bivariate function evaluation for relatively large plaintext integers has been proposed by Maeda et al. (WAHC'22). In the method of Maeda et al., the polynomial evaluation is the most time-consuming part in terms of the execution time. In this work, we construct a method for arbitrary bivariate functions based on homomorphic linear transformations, eliminating the need for polynomial evaluation to enhance efficiency. We also provide a time complexity analysis showing that our method achieves lower time computational complexity. In addition, we implement the previous method of Maeda et al. and our proposed method using OpenFHE with the input domain {0, 1,..., 2^15 - 1} and demonstrate that our method achieves a 1.38 times speed up for the division function compared to the previous method.
- 17 November 2018
- Published by Springer Nature
- p. 349-358
Privacy-Preserving Naive Bayes Classification Using Fully Homomorphic Encryption
- 17 November 2018
- Published by Springer Nature
- p. 349-358
Privacy-Preserving Naive Bayes Classification Using Fully Homomorphic Encryption
- 15 August 2018
- Published by Springer Nature
- p. 442-460
A Revocable Group Signature Scheme with Scalability from Simple Assumptions and Its Implementation
- 15 August 2018
- Published by Springer Nature
- p. 442-460
A Revocable Group Signature Scheme with Scalability from Simple Assumptions and Its Implementation
- 1 January 2018
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
- Vol. E101.A (1) , 39-50
Efficient Homomorphic Encryption with Key Rotation and Security Update
Japan's largest platform for academic e-journals: J-STAGE is a full text database for reviewed academic papers published by Japanese societies
- 1 January 2018
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
- Vol. E101.A (1) , 39-50
Efficient Homomorphic Encryption with Key Rotation and Security Update
Japan's largest platform for academic e-journals: J-STAGE is a full text database for reviewed academic papers published by Japanese societies
- Top Cited Papers
- 29 December 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 13 (5) , 1333-1345
Privacy-Preserving Deep Learning via Additively Homomorphic Encryption
We present a privacy-preserving deep learning system in which many learning participants perform neural network-based deep learning over a combined dataset of all, without revealing the participants’ local data to a central server. To that end, we revisit the previous work by Shokri and Shmatikov (ACM CCS 2015) and show that, with their method, local data information may be leaked to an honest-but-curious server. We then fix that problem by building an enhanced system with the following properties: 1) no information is leaked to the server and 2) accuracy is kept intact, compared with that of the ordinary deep learning system also over the combined dataset. Our system bridges deep learning and cryptography: we utilize asynchronous stochastic gradient descent as applied to neural networks, in combination with additively homomorphic encryption. We show that our usage of encryption adds tolerable overhead to the ordinary deep learning system.
- 29 December 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 13 (5) , 1333-1345
Top Cited PapersPrivacy-Preserving Deep Learning via Additively Homomorphic Encryption
We present a privacy-preserving deep learning system in which many learning participants perform neural network-based deep learning over a combined dataset of all, without revealing the participants’ local data to a central server. To that end, we revisit the previous work by Shokri and Shmatikov (ACM CCS 2015) and show that, with their method, local data information may be leaked to an honest-but-curious server. We then fix that problem by building an enhanced system with the following properties: 1) no information is leaked to the server and 2) accuracy is kept intact, compared with that of the ordinary deep learning system also over the combined dataset. Our system bridges deep learning and cryptography: we utilize asynchronous stochastic gradient descent as applied to neural networks, in combination with additively homomorphic encryption. We show that our usage of encryption adds tolerable overhead to the ordinary deep learning system.
- 1 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Privacy preserving extreme learning machine using additively homomorphic encryption
Recently, computational outsourcing using cloud services is getting popular for big data analysis, and many cloud sourcing providers provide machine learning platforms where we can perform various prediction and classification tasks very easily. On the other hand, there still remains a big hurdle to analyze personal big data on cloud services because the leakage of personal information is a critical issue. As a remedy for this, we propose a privacy preserving machine learning algorithm for Extreme Learning Machine (PP-ELM), which can learn from data encrypted with an additively homomorphic encryption. In the proposed outsourcing method, we consider a three-participants model consisting of data contributors, outsourced server, and data analyst. A data contributor preprocesses and encrypts data, and an outsourced server receives encrypted data and calculate hidden layer outputs using additive operations. Then, a data analyst receives the hidden outputs of ELM from the outsourced server and they are used to obtain ELM connection weights. Since the proposed outsourcing model can learn ELM over encrypted data, it is expected to mitigate a hurdle to deal with personal data on cloud services. In addition, the proposed PP-ELM allows us to learn multiple sources of personal data in a secure way, and this might lead to a better solution for a practical problem than before.
- 1 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Privacy preserving extreme learning machine using additively homomorphic encryption
Recently, computational outsourcing using cloud services is getting popular for big data analysis, and many cloud sourcing providers provide machine learning platforms where we can perform various prediction and classification tasks very easily. On the other hand, there still remains a big hurdle to analyze personal big data on cloud services because the leakage of personal information is a critical issue. As a remedy for this, we propose a privacy preserving machine learning algorithm for Extreme Learning Machine (PP-ELM), which can learn from data encrypted with an additively homomorphic encryption. In the proposed outsourcing method, we consider a three-participants model consisting of data contributors, outsourced server, and data analyst. A data contributor preprocesses and encrypts data, and an outsourced server receives encrypted data and calculate hidden layer outputs using additive operations. Then, a data analyst receives the hidden outputs of ELM from the outsourced server and they are used to obtain ELM connection weights. Since the proposed outsourcing model can learn ELM over encrypted data, it is expected to mitigate a hurdle to deal with personal data on cloud services. In addition, the proposed PP-ELM allows us to learn multiple sources of personal data in a secure way, and this might lead to a better solution for a practical problem than before.
- Open Access
- 30 October 2017
- Published by Association for Computing Machinery (ACM)
- p. 111-122
A Framework of Privacy Preserving Anomaly Detection
- 30 October 2017
- Published by Association for Computing Machinery (ACM)
- p. 111-122
Open AccessA Framework of Privacy Preserving Anomaly Detection
- 16 October 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 67 (2) , 1582-1597
Road-to-Vehicle Communications With Time-Dependent Anonymity: A Lightweight Construction and Its Experimental Results
This paper describes techniques that enable vehicles to collect local information (such as road conditions and traffic information) and report it via road-to-vehicle communications. To exclude malicious data, the collected information is signed by each vehicle. In this communications system, the location privacy of vehicles must be maintained. However, simultaneously linkable information (such as travel routes) is also important. That is, no such linkable information can be collected when full anonymity is guaranteed using cryptographic tools such as group signatures. Similarly, continuous linkability (via pseudonyms, for example) may also cause problem from the viewpoint of privacy. In this paper, we propose a road-to-vehicle communication system with relaxed anonymity via group signatures with time-token dependent linking (GS-TDL). Briefly, a vehicle is unlinkable unless it generates multiple signatures in the same time period. We provide our experimental results (using the RELIC library on a cheap and constrained computational power device, Raspberry Pi), and simulate our system by using a traffic simulator (PTV), a radio wave propagation analysis tool (RapLab), and a network simulator (QualNet). Though a similar functionality of time-token dependent linking was proposed by Wu, Domingo-Ferrer and Gonz´lez-Nicol´s (IEEE T. Vehicular Technology 2010), we can show an attack against the scheme where anyone can forge a valid group signature without using a secret key. In contrast, our GS-TDL scheme is provably secure. In addition to the time-dependent linking property, our GS-TDL scheme supports verifier-local revocation (VLR), where a signer (vehicle) is not involved in the revocation procedure. It is particularly worth noting that no secret key or certificate of a signer (vehicle) must be updated whereas the security credential management system (SCMS) must update certificates frequently for vehicle privacy. Moreover, our technique maintains constant signing and verification costs by using the linkable part of signatures. This might be of independent interest.
- 16 October 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 67 (2) , 1582-1597
Road-to-Vehicle Communications With Time-Dependent Anonymity: A Lightweight Construction and Its Experimental Results
This paper describes techniques that enable vehicles to collect local information (such as road conditions and traffic information) and report it via road-to-vehicle communications. To exclude malicious data, the collected information is signed by each vehicle. In this communications system, the location privacy of vehicles must be maintained. However, simultaneously linkable information (such as travel routes) is also important. That is, no such linkable information can be collected when full anonymity is guaranteed using cryptographic tools such as group signatures. Similarly, continuous linkability (via pseudonyms, for example) may also cause problem from the viewpoint of privacy. In this paper, we propose a road-to-vehicle communication system with relaxed anonymity via group signatures with time-token dependent linking (GS-TDL). Briefly, a vehicle is unlinkable unless it generates multiple signatures in the same time period. We provide our experimental results (using the RELIC library on a cheap and constrained computational power device, Raspberry Pi), and simulate our system by using a traffic simulator (PTV), a radio wave propagation analysis tool (RapLab), and a network simulator (QualNet). Though a similar functionality of time-token dependent linking was proposed by Wu, Domingo-Ferrer and Gonz´lez-Nicol´s (IEEE T. Vehicular Technology 2010), we can show an attack against the scheme where anyone can forge a valid group signature without using a secret key. In contrast, our GS-TDL scheme is provably secure. In addition to the time-dependent linking property, our GS-TDL scheme supports verifier-local revocation (VLR), where a signer (vehicle) is not involved in the revocation procedure. It is particularly worth noting that no secret key or certificate of a signer (vehicle) must be updated whereas the security credential management system (SCMS) must update certificates frequently for vehicle privacy. Moreover, our technique maintains constant signing and verification costs by using the linkable part of signatures. This might be of independent interest.
- 19 September 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 17 (2) , 292-305
Group Signatures with Time-Bound Keys Revisited: A New Model, an Efficient Construction, and its Implementation
Chu et al. (ASIACCS 2012) proposed group signature with time-bound keys (GS-TBK), where each signing key is associated with expiry time τ. In addition, to prove membership of the group, a signer needs to prove that the expiry time has not passed, i.e., t <; τ, where t is the current time. A signer whose expiry time has passed is automatically revoked, and this revocation is called natural revocation. Signers can be revoked simultaneously before their expiry times if the credential is compromised. This revocation is called premature revocation. A nice property in the Chu et al. proposal is that the size of revocation lists can be reduced compared to those of Verifier-Local Revocation (VLR) group signature schemes by assuming that natural revocation accounts for most of the signer revocations in practice, and prematurely revoked signers are only a small fraction. In this paper, we point out that the definition of traceability of Chu et al. did not capture the unforgeability of expiry time for signing keys, which guarantees that no adversary who has a signing key associated with expiry time τ can compute a valid signature after τ has passed. This situation significantly reduces the dependability of the system since legitimate signing keys may be used for providing a forged signature. We introduce a security model that captures unforgeability, and propose a secure GS-TBK scheme in the new model. Our scheme also provides constant signing costs, whereas those of the previous schemes depended on the bit-length of the time representation. Finally, we provide the implementation results. We employ Barreto-Lynn-Scott (BLS) curves with 455-bit prime order and the RELIC library, and demonstrate that our scheme is feasible in practical settings.
- 19 September 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 17 (2) , 292-305
Group Signatures with Time-Bound Keys Revisited: A New Model, an Efficient Construction, and its Implementation
Chu et al. (ASIACCS 2012) proposed group signature with time-bound keys (GS-TBK), where each signing key is associated with expiry time τ. In addition, to prove membership of the group, a signer needs to prove that the expiry time has not passed, i.e., t <; τ, where t is the current time. A signer whose expiry time has passed is automatically revoked, and this revocation is called natural revocation. Signers can be revoked simultaneously before their expiry times if the credential is compromised. This revocation is called premature revocation. A nice property in the Chu et al. proposal is that the size of revocation lists can be reduced compared to those of Verifier-Local Revocation (VLR) group signature schemes by assuming that natural revocation accounts for most of the signer revocations in practice, and prematurely revoked signers are only a small fraction. In this paper, we point out that the definition of traceability of Chu et al. did not capture the unforgeability of expiry time for signing keys, which guarantees that no adversary who has a signing key associated with expiry time τ can compute a valid signature after τ has passed. This situation significantly reduces the dependability of the system since legitimate signing keys may be used for providing a forged signature. We introduce a security model that captures unforgeability, and propose a secure GS-TBK scheme in the new model. Our scheme also provides constant signing costs, whereas those of the previous schemes depended on the bit-length of the time representation. Finally, we provide the implementation results. We employ Barreto-Lynn-Scott (BLS) curves with 455-bit prime order and the RELIC library, and demonstrate that our scheme is feasible in practical settings.