Electrical-level attacks on CPUs, FPGAs, and GPUs: Survey and implications in the heterogeneous era
Given the need for efficient high-performance computing, computer architectures combining
central processing units (CPUs), graphics processing units (GPUs), and field-programmable …
central processing units (CPUs), graphics processing units (GPUs), and field-programmable …
Security of neural networks from hardware perspective: A survey and beyond
Recent advances in neural networks (NNs) and their applications in deep learning
techniques have made the security aspects of NNs an important and timely topic for …
techniques have made the security aspects of NNs an important and timely topic for …
NPU thermal management
Neural processing units (NPUs) are becoming an integral part in all modern computing
systems due to their substantial role in accelerating neural networks (NNs). The significant …
systems due to their substantial role in accelerating neural networks (NNs). The significant …
BERRY: Bit Error Robustness for Energy-Efficient Reinforcement Learning-Based Autonomous Systems
Z Wan, N Chandramoorthy… - 2023 60th ACM/IEEE …, 2023 - ieeexplore.ieee.org
Autonomous systems, such as Unmanned Aerial Vehicles (UAVs), are expected to run
complex reinforcement learning (RL) models to execute fully autonomous position …
complex reinforcement learning (RL) models to execute fully autonomous position …
Stochastic-hmds: Adversarial-resilient hardware malware detectors via undervolting
MS Islam, I Alouani… - 2023 60th ACM/IEEE …, 2023 - ieeexplore.ieee.org
Machine learning-based hardware malware detectors (HMDs) offer a potential game
changing advantage in defending systems against malware. However, HMDs suffer from …
changing advantage in defending systems against malware. However, HMDs suffer from …
Exceeding conservative limits: A consolidated analysis on modern hardware margins
G Papadimitriou, A Chatzidimitriou… - … on Device and …, 2020 - ieeexplore.ieee.org
Modern large-scale computing systems (data centers, supercomputers, cloud and edge
setups and high-end cyber-physical systems) employ heterogeneous architectures that …
setups and high-end cyber-physical systems) employ heterogeneous architectures that …
Understanding power consumption and reliability of high-bandwidth memory with voltage underscaling
Modern computing devices employ High-Bandwidth Memory (HBM) to meet their memory
bandwidth requirements. An HBM-enabled device consists of multiple DRAM layers stacked …
bandwidth requirements. An HBM-enabled device consists of multiple DRAM layers stacked …
X-nvdla: Runtime accuracy configurable nvdla based on applying voltage overscaling to computing and memory units
H Afzali-Kusha, M Pedram - … on Circuits and Systems I: Regular …, 2023 - ieeexplore.ieee.org
This paper investigates a runtime accuracy reconfigurable implementation of an energy
efficient deep learning accelerator. It is based on voltage overscaling (VOS) technique which …
efficient deep learning accelerator. It is based on voltage overscaling (VOS) technique which …
An automotive case study on the limits of approximation for object detection
The accuracy of camera-based object detection (CBOD) built upon deep learning is often
evaluated against the real objects in frames only. However, such simplistic evaluation …
evaluated against the real objects in frames only. However, such simplistic evaluation …
Security of IoT device: perspective forensic/anti-forensic issues on invalid area of NAND flash memory
NAND flash memory-based IoT device can potentially still leave behind original personal
data in an invalid area even if the data has been deleted. In this paper, we raise the forensic …
data in an invalid area even if the data has been deleted. In this paper, we raise the forensic …