Convolutional neural networks (CNNs) have demonstrated significant superiority in modern artificial intelligence (AI) applications. To accelerate the inference process of CNNs, reconfigurable CNN accelerators that support diverse are widely employed for AI systems. Given ubiquitous deployment these systems, there is a growing concern regarding security and potential attacks they may face, including hardware Trojans. This paper proposes Trojan designed to attack crucial component FPGA-based...

Convolutional neural networks (CNNs) have demonstrated significant superiority in modern artificial intelligence (AI) applications. To accelerate the inference process of CNNs, reconfigurable CNN accelerators that support diverse are widely employed for AI systems. Due to ubiquitous deployment these systems, a strong incentive rises adversaries attack via hardware Trojan, which is one most important models security domain. This paper proposed Trojan attacks crucial component accelerators,...

Polynomial multiplication is a significant bottleneck for Saber. To speed it up, number theoretic transform (NTT), Toom-Cook and Schoolbook polynomial (SPM) are commonly used algorithms. Among them, SPM has no restrictions more suitable high parallel architecture. However, superior performance compatibility across different devices not simultaneously available in the current work. Therefore, we propose highly scalable Winograd-based algorithm that reduces of loops achieves 32.18% reduction...

The inference results of neural network accelerators often involve personal privacy or business secrets in intelligent systems. It is important for the safety convolutional (CNN) accelerator to prevent key data and result from being leaked. latest CNN models have started combine with fully homomorphic encryption (FHE), ensuring security. However, computational complexity, storage overhead, time are significantly increased compared traditional models. This paper proposed a lightweight FHE...

The customization of accelerators for sparse convolutional neural networks (SCNN) has been demonstrated to be a promising approach enhance the computational efficiency CNNs. However, current sparse-based works always employ reduced-scale engines (CEs), resulting in limited performance compared previous dense-based accelerators. It is found that deployment large-scale CEs encounters three challenges: inadequate utilization CEs, unstable workload and complex interconnections between...

With the boom in artificial intelligence (AI), numerous reconfigurable convolution neural network (CNN) accelerators have emerged within both industry and academia, aiming to enhance AI computing capabilities. However, this rapid landscape has also witnessed a rise hardware Trojan attacks targeted at CNN accelerators, thereby posing substantial threats reliability security of these systems. Despite escalating concern, there exists scarcity protection schemes explicitly tailored counteract...

This paper proposed a context-directed replacement policy (CRP) to improve cache hit rate for Coarse-grained reconfigurable arrays (CGRA). CRP updates the priority according usage of configuration contexts. Once finding two candidates contexts with highest priority, reserves larger context in and evict smaller one. Using five different access patterns as benchmark, experimental results show that outperforms LRU RRIP policies, especially scan mixed patterns. Under test random sequence, can...

The increasingly complicated and versatile convolutional neural networks (CNNs) models bring challenges to hardware acceleration in terms of performance, energy efficiency flexibility. This paper proposes a reconfigurable accelerator (RNA) for flexible efficient CNN acceleration. To provide flexibility, RNA employs dynamically computing framework rapidly configure data path between processing elements (PE) at run-time, as well an interlaced access mechanism multi-bank RAM. achieve high...

In this paper, the air conditioning system in pharmaceutical industry throughout life cycle energy efficiency had been evaluated according to SETAC and ISO put forward idea of "life assessment. The purpose was adopt reasonable measures saving technology minimum consumption clean minimal impact on environment load. At same time, economy reasonable. Clean wind a factory as an example. renovation effect were analyzed. CO 2 emissions analyzed its stage which provided reference for other engineering.

Coarse-grained reconfigurable arrays (CGRA) with high performance cache prefetching can mitigate off-chip memory latency. Taking advantage of the iteration number configuration context running on CGRA, this paper proposes a based (CCP) mechanism for CGRA. CCP utilizes first 256 accesses to obtain best offset demand addresses base address, and then associates it generate miss. Moreover, reduce bandwidth consumption caused by inaccurate prefetching, dynamically adjusts prefetch degree...

Artificial intelligence is changing and influencing our world. As one of the main algorithms in field artificial intelligence, convolutional neural networks (CNNs) have developed rapidly recent years. Especially after emergence NASNet, CNNs gradually pushed idea AutoML to public’s attention, large numbers new structures designed by automatic searches are appearing. These usually based on reinforcement learning evolutionary algorithms. However, sometimes, blocks these complex, there no small...

The Number Theoretic Transform (NTT) is currently widely used to accelerate polynomial multiplication in finite fields, which the computational bottleneck of lattice-based fully homomorphic encryption schemes. key challenge NTT hardware structure implementation lies large amount data storage and complex pattern access. This paper proposes an efficient access algorithm. Benefiting from specific optimization different stages NTT, use resource can be reduced by 46%~80%, performance based on...

Because of many present problems the intelligent switch in substation, it is necessary to study on components device. The hardware platform device was based high-speed low-voltage differential bus (BLVDS). By improving real-time characteristics, communication available for substation process bus. It also described that solution online condition monitoring Research these aspects laid a solid foundation