Vincent Meyers

Dr. Vincent Meyers

Karlsruhe Institute of Technology, Karlsruhe, Baden-Württemberg, Germany
orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
6 March 2023 in IEEE Design & Test
Verified
Journal Article
Closed access
Vincent MeyersVincent Meyers
Vincent Meyers
DGDennis R. E. Gnad
Dennis R. E. Gnad
MTMehdi Tahoori
Mehdi Tahoori
This article emphasizes the growing importance of studying defenses for physical attacks on neural network accelerators, and describes approaches based on side-channel attacks and fault-injection attacks.

28 September 2025
Verified
Conference Paper
Open access
MTMehdi Tahoori
Mehdi Tahoori
Vincent MeyersVincent Meyers
Vincent Meyers
MRMahboobe Sadeghipour Roodsari
Mahboobe Sadeghipour Roodsari
HXHuashuangyang Xu
Huashuangyang Xu
JBJuergen Becker
Juergen Becker
MAMohamed Shelkamy Ali
Mohamed Shelkamy Ali

1 January 2026 in IEEE Transactions on Device and Materials Reliability
Verified
Journal Article
Closed access
MRMahboobe Sadeghipour Roodsari
Mahboobe Sadeghipour Roodsari
Vincent MeyersVincent Meyers
Vincent Meyers
MTMehdi Tahoori
Mehdi Tahoori

7 January 2025 in Cryptography
Verified
Journal Article
Open access
PDF file
Vincent MeyersVincent Meyers
Vincent Meyers
MHMichael Hefenbrock
Michael Hefenbrock
DGDennis R. E. Gnad
Dennis R. E. Gnad
MTMehdi Tahoori
Mehdi Tahoori
Neural networks have become pivotal in advancing applications across various domains, including healthcare, finance, surveillance, and autonomous systems. To achieve low latency and high efficiency, field-programmable gate arrays (FPGAs) are increasingly being employed as accelerators for neural network inference in cloud and edge devices. However, the rising costs and complexity of neural network training have led to the widespread use of outsourcing of training, pre-trained models, and machine learning services, raising significant concerns about security and trust. Specifically, malicious actors may embed neural Trojans within NNs, exploiting them to leak sensitive data through side-channel analysis. This paper builds upon our prior work, where we demonstrated the feasibility of embedding Trojan side-channels in neural network weights, enabling the extraction of classification results via remote power side-channel attacks. In this expanded study, we introduced a broader range of experiments to evaluate the robustness and effectiveness of this attack vector. We detail a novel training methodology that enhanced the correlation between power consumption and network output, achieving up to a 33% improvement in reconstruction accuracy over benign models. Our approach eliminates the need for additional hardware, making it stealthier and more resistant to conventional hardware Trojan detection methods. We provide comprehensive analyses of attack scenarios in both controlled and variable environmental conditions, demonstrating the scalability and adaptability of our technique across diverse neural network architectures, such as MLPs and CNNs. Additionally, we explore countermeasures and discuss their implications for the design of secure neural network accelerators. To the best of our knowledge, this work is the first to present a passive output recovery attack on neural network accelerators, without explicit trigger mechanisms. The findings emphasize the urgent need to integrate hardware-aware security protocols in the development and deployment of neural network accelerators.

20 January 2025
Verified
Conference Paper
Open access
Vincent MeyersVincent Meyers
Vincent Meyers
MHMichael Hefenbrock
Michael Hefenbrock
MSMahboobe Sadeghipourrudsari
Mahboobe Sadeghipourrudsari
DGDennis R. E. Gnad
Dennis R. E. Gnad
MTMehdi Tahoori
Mehdi Tahoori
For AI hardware, functional safety is crucial, especially for neural network (NN) accelerators used in safety-critical systems. A key requirement for maintaining this safety is the precise detection of out-of-distribution (OOD) instances, which are inputs significantly distinct from the training data. Neglecting to integrate robust OOD detection may result in possible safety hazards, diminished performance, and inaccurate decision-making within NN applications. Existing methods for OOD detection have been explored for full-precision models. However, the evaluation of methods on quantized neural network (QNN), which are often deployed on hardware accelerators such as FPGAs, and on-device hardware realization of concurrent OOD detection (COD) is missing in literature. In this paper, we provide a novel approach to OOD detection for NN FPGA accelerators using power measurements. Utilizing the power side-channel through digital voltage sensors allows on-device OOD detection in a non-intrusive and concurrent manner, without relying on explicit labels or modifications to the underlying NN. Furthermore, our method allows OOD detection before the inference finishes. Additionally to the evaluation, we provide an efficient hardware implementation of COD on an actual FPGA.

16 January 2026 in IACR Transactions on Cryptographic Hardware and Embedded Systems
Verified
Journal Article
Open access
PDF file
HXHuashuangyang Xu
Huashuangyang Xu
SMSergej Meschkov
Sergej Meschkov
Vincent MeyersVincent Meyers
Vincent Meyers
MTMehdi Tahoori
Mehdi Tahoori
The explosive growth of AI-driven services has led to cloud-based Field Programmable Gate Array (FPGA) accelerators as key enablers of high-performance training and inference in modern data centers. Since 2024, the demand for deploying large AI workloads, especially Large Language Model (LLM), in the cloud has increased dramatically, intensifying competition among cloud providers and increasing pressure on shared FPGA infrastructures. This increasing reliance highlights the need for robust hardware security measures for cloud FPGAs. A particularly serious threat is fault injection attacks, which exploit dynamic voltage fluctuations to induce timing faults, potentially compromising functional integrity and bypassing cryptographic protections. However, existing verification procedures and structural Design Rule Check (DRC) remain blind to attacks embedded in benign-looking circuits. In this paper, we present Power-Wasting Neural Network (PWNN), a novel adversarial technique that leverages the inherent switching behavior of neural network operations to act as a power-waster circuit under adversarial input patterns. We systematically explore network architectures, and input patterns to craft configurations that induce voltage fluctuations capable of triggering timing faults for successful Differential Fault Analysis (DFA). Our PWNN implementation uses a standard open-source tool chain and passes all pre-implementation verification checks, while covertly inducing faults at runtime. We demonstrate on both the AMD ZCU104 and PYNQ-Z2 that PWNN can reliably cause timing faults on the critical path of a co-located AES-128 block cipher, enabling the rapid collection of correct/faulty ciphertext pairs needed for DFA-based key recovery. These results show that functionally correct, DRC compliant accelerators can serve as powerful, adaptive fault injectors that invalidate assumptions about bitstream security and hardware isolation.