Keiji Kimura
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Keiji Kimura has a total of 73 co-authors. Here are the top co-authors by number of shared publications:
- Hironori Kasahara
Waseda University
9 shared publications - Hiroki Mikami
Waseda University
4 shared publications - Tohma Kawasumi
Waseda University
4 shared publications - Ahmed El-Mahdy
Computer Science and Engineering Department, Egypt-Japan University of Science and Technology, Alexandria, Egypt
3 shared publications Hiroki Mikami
Waseda University
2 shared publicationsMostafa I. Soliman
Department of Computer Science and Engineering, New Borg El-Arab City, EgyptJapan University of Science and Technology, Alexandria, Egypt
2 shared publications
Latest Publications
- Open Access
- 1 June 2026
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Information and Systems
- Vol. E109.D (6) , 894-910
Efficient Memory Protection Method for Large-Scale Host-Enclave Data Transfer on Keystone Enclave
Enclave-type Trusted Execution Environments (TEEs) provide a hardware-isolated environment, called an enclave, where confidential applications can be executed securely. The enclave runtime is designed to be fully trusted but tends to have limited functions due to its simple implementation, which prioritizes security. Thus, it relies on untrusted host OS functions, particularly I/O operations, through a minimum secure communication interface between a host and an enclave. Such a communication interface across the boundary between trusted and untrusted domains must be implemented in a manner that does not compromise security. However, tight security constraints can lead to a loss of performance and compatibility for applications. Achieving efficient and flexible secure communication is a challenging issue. Keystone Enclave is one of the representative enclave-type TEE implementations for RISC-V. While Keystone equips a set of edge calls as a communication interface, it introduces data transfer efficiency issues and security concerns. When transferring large amounts of data from a host to an enclave, the edge calls introduce a significant data transfer overhead due to the restrictions on Keystone's implementation of memory isolation. Besides, the original edge calls do not protect the transferred data from other programs. This paper proposes a secure, efficient, and scalable host-enclave data transfer method for Keystone Enclave. The proposed method introduces an additional memory region dedicated to the protection of data transfer. This region is protected independently of the enclaves and other shared memory. It is enforced that the contents in the region are validated before use by privileged software. This approach enables efficient and scalable data transfer, as well as flexible data protection, without requiring additional hardware extensions. The evaluation of the proposed method on the HiFive Unmatched RISC-V board shows 2.2-2.4× better performance than the method using the original edge calls for large-size data transfer. We also evaluate the performance of the I/O system call delegation using the proposed method and confirm its practicality.
- 1 June 2026
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Information and Systems
- Vol. E109.D (6) , 894-910
Open AccessEfficient Memory Protection Method for Large-Scale Host-Enclave Data Transfer on Keystone Enclave
Enclave-type Trusted Execution Environments (TEEs) provide a hardware-isolated environment, called an enclave, where confidential applications can be executed securely. The enclave runtime is designed to be fully trusted but tends to have limited functions due to its simple implementation, which prioritizes security. Thus, it relies on untrusted host OS functions, particularly I/O operations, through a minimum secure communication interface between a host and an enclave. Such a communication interface across the boundary between trusted and untrusted domains must be implemented in a manner that does not compromise security. However, tight security constraints can lead to a loss of performance and compatibility for applications. Achieving efficient and flexible secure communication is a challenging issue. Keystone Enclave is one of the representative enclave-type TEE implementations for RISC-V. While Keystone equips a set of edge calls as a communication interface, it introduces data transfer efficiency issues and security concerns. When transferring large amounts of data from a host to an enclave, the edge calls introduce a significant data transfer overhead due to the restrictions on Keystone's implementation of memory isolation. Besides, the original edge calls do not protect the transferred data from other programs. This paper proposes a secure, efficient, and scalable host-enclave data transfer method for Keystone Enclave. The proposed method introduces an additional memory region dedicated to the protection of data transfer. This region is protected independently of the enclaves and other shared memory. It is enforced that the contents in the region are validated before use by privileged software. This approach enables efficient and scalable data transfer, as well as flexible data protection, without requiring additional hardware extensions. The evaluation of the proposed method on the HiFive Unmatched RISC-V board shows 2.2-2.4× better performance than the method using the original edge calls for large-size data transfer. We also evaluate the performance of the I/O system call delegation using the proposed method and confirm its practicality.
- 28 June 2025
- Published by Springer Nature
- p. 34-49
Evaluation of Speedup and Energy with Multigrain Parallelizing Compiler
- 28 June 2025
- Published by Springer Nature
- p. 34-49
Evaluation of Speedup and Energy with Multigrain Parallelizing Compiler
- 18 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 18, 568-575
Parallel Verification in RISC-V Secure Boot
Assuring the reliability of the OS boot process is essential to realize reliable computer systems. Secure Boot enables it by introducing the verification of the boot image with digital signature and hash values. This can be the basis for various security mechanisms. However, Secure Boot requires a long boot time due to their expensive computation costs, resulting in extended downtime. In this paper, we first implement Secure Boot in an ordinary RISC-V boot process on U-Boot, a representative open-source bootloader, and clarify the overhead introduced by the verification process. Based on the insight obtained above, we propose a parallelization of the verification process on a multi-core in Secure Boot. It accelerates the boot process while securely authenticating the boot image. We implement the proposed parallel verification process on U-Boot. The evaluation on a HiFive Unmatched RISC-V board shows that the parallelized hash computation on four cores achieves 3.96 times better performance than the original.
- 18 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 18, 568-575
Parallel Verification in RISC-V Secure Boot
Assuring the reliability of the OS boot process is essential to realize reliable computer systems. Secure Boot enables it by introducing the verification of the boot image with digital signature and hash values. This can be the basis for various security mechanisms. However, Secure Boot requires a long boot time due to their expensive computation costs, resulting in extended downtime. In this paper, we first implement Secure Boot in an ordinary RISC-V boot process on U-Boot, a representative open-source bootloader, and clarify the overhead introduced by the verification process. Based on the insight obtained above, we propose a parallelization of the verification process on a multi-core in Secure Boot. It accelerates the boot process while securely authenticating the boot image. We implement the proposed parallel verification process on U-Boot. The evaluation on a HiFive Unmatched RISC-V board shows that the parallelized hash computation on four cores achieves 3.96 times better performance than the original.
- Open Access
- 23 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 119342-119351
Lightweight Histological Tumor Classification Using a Joint Sparsity-Quantization Aware Training Framework
Cancer decision-making is a complex process that can be exacerbated by the limited availability of oncological expertise. This is particularly true in rural areas and settings with fewer resources. Recently, there has been an interest in the potential of artificial intelligence in reliable computer-aided diagnosis tools in such settings. Nevertheless, the majority of deep learning algorithms are resource hungry in terms of data and storage requirements. In this work, we propose a novel lightweight deep learning model for histological tumor classification through a Joint Sparsity-Quantization Aware Training framework. Extensive experiments were conducted to evaluate the proposed framework. This work aims at opening doors toward efficient point-of-care diagnostic devices suitable for environments with limited resources.
- 23 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 119342-119351
Open AccessLightweight Histological Tumor Classification Using a Joint Sparsity-Quantization Aware Training Framework
Cancer decision-making is a complex process that can be exacerbated by the limited availability of oncological expertise. This is particularly true in rural areas and settings with fewer resources. Recently, there has been an interest in the potential of artificial intelligence in reliable computer-aided diagnosis tools in such settings. Nevertheless, the majority of deep learning algorithms are resource hungry in terms of data and storage requirements. In this work, we propose a novel lightweight deep learning model for histological tumor classification through a Joint Sparsity-Quantization Aware Training framework. Extensive experiments were conducted to evaluate the proposed framework. This work aims at opening doors toward efficient point-of-care diagnostic devices suitable for environments with limited resources.
- 10 May 2023
- Published by Springer Nature
- p. 123-138
Parallelizing Factory Automation Ladder Programs by OSCAR Automatic Parallelizing Compiler
- 10 May 2023
- Published by Springer Nature
- p. 123-138
Parallelizing Factory Automation Ladder Programs by OSCAR Automatic Parallelizing Compiler
- 10 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 21 (2) , 77-80
Open-Source Hardware Memory Protection Engine Integrated With NVMM Simulator
With growing on-device IoT processing, security on edge devices becomes increasingly important. Among Trusted Execution Environment (TEE), an open-source RISC-V Keystone TEE is the expected one. However, some issues remain when applying it to various devices: untrusted DRAM, and untrusted path to non-volatile storage. These issues can be resolved by Memory Protection Engine (MPE) based on an integrity tree, and Non-Volatile Main Memory (NVMM), respectively. TEE, MPE, and NVMM must be cooperatively optimized to exploit performance. Despite this demand, there is no widely available platform which enables fast, reliable, and system-wide evaluation. In the paper, we provide an open-source hardware simulator for secure edge devices. We implemented an MPE using SGX-style Integrity Tree on the Keystone compatible RISC-V SoC. Then, we ported the NVMM simulation technique to it. Its whole design was publicized to widely provide a baseline hardware design. The MPE behavior was validated by using micro benchmarks. It revealed that the MPE read/write overhead is$2.55\times /4.16\times$ on DRAM, and$3.05\times /5.40\times$ on NVMM, respectively. Besides, we discuss our work's role by comparing with the gem5 considering TEE evaluation time and impact of the protected NVMM.- 10 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 21 (2) , 77-80
Open-Source Hardware Memory Protection Engine Integrated With NVMM Simulator
With growing on-device IoT processing, security on edge devices becomes increasingly important. Among Trusted Execution Environment (TEE), an open-source RISC-V Keystone TEE is the expected one. However, some issues remain when applying it to various devices: untrusted DRAM, and untrusted path to non-volatile storage. These issues can be resolved by Memory Protection Engine (MPE) based on an integrity tree, and Non-Volatile Main Memory (NVMM), respectively. TEE, MPE, and NVMM must be cooperatively optimized to exploit performance. Despite this demand, there is no widely available platform which enables fast, reliable, and system-wide evaluation. In the paper, we provide an open-source hardware simulator for secure edge devices. We implemented an MPE using SGX-style Integrity Tree on the Keystone compatible RISC-V SoC. Then, we ported the NVMM simulation technique to it. Its whole design was publicized to widely provide a baseline hardware design. The MPE behavior was validated by using micro benchmarks. It revealed that the MPE read/write overhead is$2.55\times /4.16\times$ on DRAM, and$3.05\times /5.40\times$ on NVMM, respectively. Besides, we discuss our work's role by comparing with the gem5 considering TEE evaluation time and impact of the protected NVMM.- 20 April 2022
- Published by Elsevier in Journal of Parallel and Distributed Computing
- Vol. 166, 57-70
Data stream clustering for low-cost machines
- 20 April 2022
- Published by Elsevier in Journal of Parallel and Distributed Computing
- Vol. 166, 57-70
Data stream clustering for low-cost machines
- 24 March 2022
- Published by Springer Nature
- p. 50-64
Performance Evaluation of OSCAR Multi-target Automatic Parallelizing Compiler on Intel, AMD, Arm and RISC-V Multicores
- 24 March 2022
- Published by Springer Nature
- p. 50-64
Performance Evaluation of OSCAR Multi-target Automatic Parallelizing Compiler on Intel, AMD, Arm and RISC-V Multicores
- Open Access
- 21 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 31626-31640
Accelerating Data Dependence Profiling Through Abstract Interpretation of Loop Instructions
Data dependence analysis is a must-do operation for parallelisation since it reveals the safe parallelisable regions of serial codes. Generally, it relies on dynamic analysis, which incurs substantial execution time and memory space overheads. As a result, there have been many efforts in the literature to strike a balance between accuracy and runtime overhead. The approaches generally rely on random instruction sampling, parallelising analysis, as well as filtering statically determined dependencies and independencies. This paper considers an alternate approach of conducting static analysis at runtime, exploiting available states just before executing loops, potentially improving precision. In particular, the paper adopts abstract interpretation using interval, congruent, and bisector domains for detecting memory data dependencies in binary programs at runtime. Abstract interpretation has the advantage of being associated with the execution semantics, making it more natural to model binary instruction execution. The profiler is implemented on top of the Pin framework and evaluated using the Polyhedral, NPB, and SPEC 2006 benchmarks suites. Results show a mean accuracy of 90.4% with an average $16.3 \times$ speedup in time in comparison with related work, making it a promising approach.
- 21 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 31626-31640
Open AccessAccelerating Data Dependence Profiling Through Abstract Interpretation of Loop Instructions
Data dependence analysis is a must-do operation for parallelisation since it reveals the safe parallelisable regions of serial codes. Generally, it relies on dynamic analysis, which incurs substantial execution time and memory space overheads. As a result, there have been many efforts in the literature to strike a balance between accuracy and runtime overhead. The approaches generally rely on random instruction sampling, parallelising analysis, as well as filtering statically determined dependencies and independencies. This paper considers an alternate approach of conducting static analysis at runtime, exploiting available states just before executing loops, potentially improving precision. In particular, the paper adopts abstract interpretation using interval, congruent, and bisector domains for detecting memory data dependencies in binary programs at runtime. Abstract interpretation has the advantage of being associated with the execution semantics, making it more natural to model binary instruction execution. The profiler is implemented on top of the Pin framework and evaluated using the Polyhedral, NPB, and SPEC 2006 benchmarks suites. Results show a mean accuracy of 90.4% with an average $16.3 \times$ speedup in time in comparison with related work, making it a promising approach.