TY - JOUR
T1 - A Generic Cryptographic Deep-Learning Inference Platform for Remote Sensing Scenes
AU - Chen, Qian
AU - Wu, Yulin
AU - Wang, Xuan
AU - Jiang, Zoe L.
AU - Zhang, Weizhe
AU - Liu, Yang
AU - Alazab, Mamoun
N1 - Publisher Copyright:
© 2008-2012 IEEE.
PY - 2023
Y1 - 2023
N2 - Deep learning plays an essential role in multidisciplinary research of remote sensing. We will encounter security problems during the data acquisition, processing, and result generation stages. Therefore, secure deep-learning inference services are one of the most important links. Some theoretical progress has been made in cryptographic deep-learning inference, but it lacks a general platform that can be realized in reality. Constantly modifying the corresponding models to approximate the plaintext results reveal the model information to a certain extent. This article proposes a generic post-quantum platform named the PyHENet, which perfectly combines cryptography with plaintext deep learning libraries. Second, we optimize the convolution, activation, and pooling functions and complete the ciphertext operation under floating point numbers for the first time. Moreover, the computation process is accelerated by single instruction multiple data streams and GPU parallel computing. The experimental results show that the PyHENet is closer to the plaintext inference platform than any other cryptographic model and has satisfactory robustness. The optimized PyHENet obtained a better accuracy of 95.05% in the high-resolution NaSC-TG2 database, which the Tiangong-2 space station received.
AB - Deep learning plays an essential role in multidisciplinary research of remote sensing. We will encounter security problems during the data acquisition, processing, and result generation stages. Therefore, secure deep-learning inference services are one of the most important links. Some theoretical progress has been made in cryptographic deep-learning inference, but it lacks a general platform that can be realized in reality. Constantly modifying the corresponding models to approximate the plaintext results reveal the model information to a certain extent. This article proposes a generic post-quantum platform named the PyHENet, which perfectly combines cryptography with plaintext deep learning libraries. Second, we optimize the convolution, activation, and pooling functions and complete the ciphertext operation under floating point numbers for the first time. Moreover, the computation process is accelerated by single instruction multiple data streams and GPU parallel computing. The experimental results show that the PyHENet is closer to the plaintext inference platform than any other cryptographic model and has satisfactory robustness. The optimized PyHENet obtained a better accuracy of 95.05% in the high-resolution NaSC-TG2 database, which the Tiangong-2 space station received.
KW - Convolutional neural network
KW - Cryptography
KW - Deep learning
KW - deep learning inference
KW - fully homomorphic encryption
KW - Homomorphic encryption
KW - privacy preserving
KW - Remote sensing
KW - remote sensing scenes
KW - Resists
KW - Security
KW - Task analysis
KW - Convolutional neural network (CNN)
KW - fully homomorphic encryption (HE)
UR - http://www.scopus.com/inward/record.url?scp=85151543296&partnerID=8YFLogxK
U2 - 10.1109/JSTARS.2023.3260867
DO - 10.1109/JSTARS.2023.3260867
M3 - Article
AN - SCOPUS:85151543296
SN - 1939-1404
VL - 16
SP - 3309
EP - 3321
JO - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
JF - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
ER -