Munkhbaatar Chinbat

Mr. Munkhbaatar Chinbat

School of Integrated Circuits, Tsinghua University, Beijing, China
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 April 2024 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Verified
Journal Article
Closed access
LWLe Wu
Le Wu
LWLiji Wu
Liji Wu
XZXiangmin Zhang
Xiangmin Zhang
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
Deep neural network (DNN) accelerators are widely used in cloud-edge-end and other application scenarios. Researchers recently focused on extracting secret information from DNN through side-channel attacks (SCAs), which substantially threaten AI security. In this brief, we propose a high-security, high-performance side-channel countermeasure using dual-rail precharge logic (DPL) for the DNN systolic array. By collecting and analyzing 5000 power traces, our proposed DPL-based systolic array provides a significantly lower correlation coefficient of 0.045. Through system-level side-channel security evaluation on field-programmable gate arrays (FPGAs), the DPL-based systolic array can effectively defend against weight extraction under power SCAs.

9 October 2025 in Cryptography
Verified
Journal Article
Open access
PDF file
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
LWLiji Wu
Liji Wu
XZXiangmin Zhang
Xiangmin Zhang
YYYifan Yang
Yifan Yang
MWMan Wei
Man Wei
Deep learning-based side-channel analysis is one of the most effective techniques for extracting and classifying sensitive information from a target device. This paper demonstrates the best-performing deep learning model for the target implementation by evaluating various deep learning architectures, including MLP, CNN, and RNN, while systematically optimizing their hyperparameters to achieve the best performance. The paper uses a case study of the Number Theoretic Transform accelerator for the CRYSTALS-Kyber key encapsulation mechanism to show that enhanced deep learning analysis can be used to break security. The best-performing deep learning-based model achieved a 96.64% accuracy in classifying pairwise coefficients of the s vector, which is used to generate the secret key with the NTT accelerator for Kyber768 and Kyber1024. For Kyber512, the model achieved an accuracy of 95.71%. The proposed approach significantly improves average training efficiency, with POIs achieving up to 1.45 times faster performance for MLP models, 10.53 times faster for CNNs, and 10.28 times faster for RNNs compared to deep learning methods without POIs, while maintaining high accuracy in side-channel analysis.

30 October 2020
Verified
Conference Paper
Closed access
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
LWLiji Wu
Liji Wu
ABAltantsooj Batsukh
Altantsooj Batsukh
UKUyangaa Khuchit
Uyangaa Khuchit
XZXiangmin Zhang
Xiangmin Zhang
WYWei Yang
Wei Yang
An efficient implementation of the multiplication part is one of the significant procedures of the cryptography algorithms. In this paper, the six altered parallel multiplication methods are proposed to implement in 192-bit for the SM2 algorithm. The CPAM, CSAM, Tri-Section Pezaris, Baugh-Wooley array, Modified Booth, and the Montgomery multipliers are compared by considering minimum operational speed, area, and power. We used a mod m reducer circuit for comparing with similar outputs of the multiplier architectures. Through the final comparison, the Montgomery gives the efficient result by 504 LUTs, 5.532ns timing, 0.101mW dynamic power. The proposed work is implemented on the Xilinx Virtex-7 FPGA board, and the programming language is VHDL.

1 December 2023
Verified
Conference Paper
Closed access
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
LWLiji Wu
Liji Wu
XZXiangmin Zhang
Xiangmin Zhang
ABAltantsooj Batsukh
Altantsooj Batsukh
YYYifan Yang
Yifan Yang
LWLe Wu
Le Wu
CRYSTALS-Kyber is a new algorithm that the NIST recently selected to standardize public-key encryption and key establishment. Therefore, studies are needed to evaluate the side-channel attack resistance of CRYSTALS-Kyber implementations. This paper presents a simple power analysis of a CRYSTALS-Kyber hardware implementation with the security parameter k=3. Since hardware implementations perform computations in parallel, the power consumption of each operation is difficult to quantify. The entire power consumption trace was identified using 6,072,500 samples during the CCAKEM implementation of Kyber. A significant part of the message encoding power consumption occurred during decapsulation. These findings show that existing hardware implementations of CRYSTALS-Kyber require effective countermeasures to efficiently resist side-channel attacks.

1 November 2024
Verified
Conference Paper
Closed access
NWNuo Wang
Nuo Wang
LWLiji Wu
Liji Wu
LLLei Li
Lei Li
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
XZXiangmin Zhang
Xiangmin Zhang
RISC-V, as a new generation open-source Reduced Instruction Set Computing (RISC) architecture, has the advantages of low power consumption, small footprint, and high performance. Especially, RISC-V processors realized based on FPGA can be applied in different fields such as embedded systems, Artificial Intelligence (AI) chips, security devices and high-performance computing. It is foreseeable that a breakthrough in quantum computers will jeopardize all communications. Therefore, processors must integrate secure quantum computer anti-encryption techniques. Currently, the security and reliability aspects of processor implementations are still greatly underexplored. To address this gap, we propose three contributions. First, we augment the open-source IP Hummingbird E203 with a hardware module for implementing the Number-Theoretic Transform (NTT) operations of the Post-Quantum Cryptography (kyber). Second, we employ the hardware/software co-design techniques to accelerate NTT operations. Third, analyze the results of implementation. Prototyping was performed on FPGA to evaluate our designed processor NTT extension instructions. The cycle count for NTT operations is reduced by 84.55% and the instruction count is reduced by 81.25%. Enhanced computational efficiency applicable to information security systems under potential quantum computing threats is shown. The scalability of Post-Quantum Cryptography (kyber) based on riscv is realized.

1 November 2024
Verified
Conference Paper
Closed access
YYYifan Yang
Yifan Yang
LWLiji Wu
Liji Wu
XZXiangming Zhang
Xiangming Zhang
MCMunkhbaatar Chinbat
Munkhbaatar Chinbat
CRYSTALS-Kyber is a post-quantum cryptography that can effectively resist quantum attacks. It has been included in the draft standard by NIST. Therefore, it is greatly important to research the side-channel security of its hardware implementation. In this paper, we propose an optimized chosen-ciphertext power analysis method for the hardware implementation of the CRYSTALS-Kyber. We conducted the side-channel security analysis experiments using the CW305 Artix-7 FPGA and analyzed the experimental results. The results demonstrate that our method can effectively exploit the side-channel leakage information from the hardware implementation of CRYSTALS-Kyber, improving the accuracy of the attack, reducing the number of power traces required, and achieving a key recovery accuracy of 98.8%.