Fast and Light-weight Binarized Neural Network Implemented in an FPGA using LUT-based Signal Processing and its Time-domain Extension for Multi-bit Processing

Fuchikami, Ryuji; Issiki, F.

doi:10.1109/icce-berlin47944.2019.8966187

Cited by 12 publications

(8 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A number of implementations for AI devices on FPGA were so far proposed 2–6 ; there are mainly two approaches. With the first approach, input to Neural Network (NN) and summation of products of weights are realized using a multiplier.…”

Section: Preliminarymentioning

confidence: 99%

“…A number of implementations for AI devices on FPGA were so far proposed [2][3][4][5][6] ; there are mainly two approaches.…”

Section: Ai Devicesmentioning

confidence: 99%

“…[2][3][4] The other approach involves implementation using LUT, as in this study. LUTnetwork 5,6 is known as a typical implementation of this kind.…”

Section: Ai Devicesmentioning

confidence: 99%

“…A number of implementation technologies are studied toward realization of such edge AI. [2][3][4][5][6] A white paper on AI 7 points out that promotion of edge IT use involves issues of safety and security risks. AI information theft attacks, malfunction attacks, and other threats were reported so far.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Hierarchical hardware trojan for LUT‐based AI devices and its evaluation

Nozaki

Takemoto

Ikezaki

et al. 2022

Elect Comm in Japan

View full text Add to dashboard Cite

To realize Society 5.0, edge AI techniques have attracted attention. On the other hand, security issues of edge AI have been reported. In addition, in the field of hardware security, the threat of hardware Trojan (HT) is emphasized. To defend the AI device from malicious attacks, it is important to check the vulnerability against various attacks. Therefore, this study proposes a new HT for AI inference devices. The proposed HT falsifies the inference result with respect to an arbitrary trigger input. The proposed HT concentrates on the Lookup Table (LUT) structure, and can be achieved by rewriting the LUT table information. As a result, the proposed HT does not need additional trojan trigger and payload circuits, that is, it can be implemented without the circuit overhead. Experiments by field programable gate array show the validity of the proposed HT.

show abstract

Section: Preliminarymentioning

confidence: 99%

“…A number of implementations for AI devices on FPGA were so far proposed [2][3][4][5][6] ; there are mainly two approaches.…”

Section: Ai Devicesmentioning

confidence: 99%

“…[2][3][4] The other approach involves implementation using LUT, as in this study. LUTnetwork 5,6 is known as a typical implementation of this kind.…”

Section: Ai Devicesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Hierarchical hardware trojan for LUT‐based AI devices and its evaluation

Nozaki

Takemoto

Ikezaki

et al. 2022

Elect Comm in Japan

View full text Add to dashboard Cite

show abstract

“…Therefore, one can assume that the proposed method can be applied universally as long as two redundant NNs are realized via combinatorial circuits. There are a number of methods for combinatorial circuit implementation of NN; particularly, one can think of methods based on the NN implementation in this study or LUT‐network reported in literature 20 …”

Section: Evaluation Experimentsmentioning

confidence: 99%

AI hardware oriented neural network physical unclonable function and its evaluation

Nozaki

Shibagaki

Takemoto

et al. 2020

Elect Comm in Japan

View full text Add to dashboard Cite

AI techniques are required for realizing society 5.0. For the issues of societal implementation for AI, an AI hardware, which is enhanced security performances including authentication, and so on, is needed in order to reduce security risks. This study proposes a new physical unclonable function (PUF) based on neural network (NN) called NN PUF. The proposed NN PUF uses a difference of calculation time in NN due to production variations of semiconductor. Evaluation experiments using a field programmable gate array (FPGA) prove the effectiveness of the proposed NN PUF.

show abstract