LSU ECE, CS Professors Win Best Paper for Work on Deep Learning Neural Networks
November 5, 2019
BATON ROUGE, LA – LSU Electrical and Computer Engineering Professor Lu Peng and LSU Computer Science and Engineering Associate Professor Jian Zhang recently received the Best Paper Award at the 10th International Green and Sustainable Computing Conference for their paper, “Hardware Accelerator for Adversarial Attacks on Deep Learning Neural Networks.”
Peng and Zhang co-authored the paper with ECE PhD students Haoqiang Guo and Fang Qi, as well as 2011 ECE PhD graduate Lide Duan.
Recent studies have identified that deep-learning neural networks, or DNNs, are vulnerable to subtle deviations, which are not perceptible to human visual systems but can fool the DNN models and lead to wrong outputs. A class of adversarial attack network algorithms has been proposed to generate robust physical deviations under different circumstances. These algorithms are the first efforts to move forward secure deep learning by providing an avenue to train future defense networks. However, their intrinsic complexity prevents their broader usage.
In the group’s paper, they propose the first hardware accelerator for adversarial attacks based on memristor crossbar arrays, which provide a power-efficient solution for information storage and processing. Their design significantly improves the throughput of a visual adversarial perturbation system, which can further improve the robustness and security of future deep learning systems. Based on the algorithm’s uniqueness, the group proposes four implementations for the adversarial attack accelerator to improve throughput, energy efficiency, and computational efficiency.
Contact: Joshua Duplechain
Director of Communications