Convolutional Neural Network Framework for Encrypted Image Classification in Cloud-Based ITS

Abstract

Internet of Things (IoT) and Cloud Computing (CC) technologies are becoming critical requirements to the advancement of intelligent transportation systems (ITSs). ITSs generally rely on captured images to evaluate the status of traffic and perform vehicle statistics. However, such images may contain confidential information, and thus, securing such images is paramount. Therefore, we propose in this paper an efficient framework for improving the security of CC-IoT based ITSs. The proposed framework allows extracting particular vehicle information without revealing any sensitive information. Towards this goal, a convolutional neural network is used to classify encrypted images, based on the vehicle type in real-time, obtained by cameras integrated into road-side units that are part of an ITS leaving sensitive information in all images hidden. Within the proposed framework, we develop a new image classification architecture that never fully decrypts the captured images, thus protecting drivers' personal information, such as location, license plate, and vehicle contents. In addition, the system does not require a fully decrypted image, which increases the system computational efficiency as compared to conventional systems. The obtained results show that the proposed partial decryption classification technique presents up to 18% reduction in average computational complexity when compared with a fully decrypted system.

Privacy-Preserving Image Classification With Deep Learning and Double Random Phase Encoding

“…Several studies have evaluated information retrieval and classification using encrypted data stored in the cloud to protect sensitive information. In [17], images were captured using roadside units, and the vehicles in those images were segmented through edge detection. Furthermore, segmented images are encrypted using a suitable algorithm using a selected mode of operation, and the encrypted data were classified based on the convolutional neural networks.…”

“…In addition, decryption processing within the cloud will reveal original images to unauthorized parties, such as cloud provider companies. Consequently, data processing based on encrypted information is promising and has become critical [15][16][17][18][19][20][21][22][23][24]. Image data are typically large, and image encryption with methods, such as data encryption standard (DES) or advanced encryption standard (AES), may be timeconsuming and unsuitable for low-computing devices, with limited power and computation capability, in IoT systems.…”

Trusted Computation Using ABM and PBM Decision Models for ITS

“…The approach proposed by Kang et al [52], which used the same dataset as mentioned above, is used as the baseline approach. Lidkea et al [20] proposed an improved ITS framework, where the information regarding a vehicle was extracted without revealing private data. The authors used an encryption algorithm in convolution neural networks to access real-time information from the vehicles.…”

“…The results indicated an 18% reduction in computational complexity for decrypting half images. We compared the proposed mechanism against the approaches of Kang et al [52] and Lidkea et al [20].…”

“…In addition, an intruder may launch a data falsification attack, also known as a false data injection attack, to perform various malicious activities (active/passive threats) inside the network [17]- [19]. Lidkea et al [20] have identified another critical threat specific to ITSs using social media, where attackers gain access to sensitive information through e-mails and other social media interactions [21]- [24] by stealing the database or records of a device or a server.…”