Improved encrypted AI robot for package recognition in IoT logistics environment

Physical security deals with package recognition for video surveillance. Robots play an important role in understanding the dynamic recognition of package images and their movement. Recognition of package images in robotic applications at high speed is a necessity. The in-built programming function in the onboard sources of robotics limits the computation speed of the recognition of the package images. By introducing the cloud environment into the package recognition, this limitation will be overcome. This cloud-based technology has an improved computation speed, high memory capacity, less cost, and flexibility in throughput. Therefore, we can store the package images in the cloud environment. Users can access the image from the cloud storage; however, the transmission of the image may create a lack of security and privacy attacks on the package image database. To overcome this security issue and maintain the privacy of the package, image database encryption is needed. We propose a secure encrypted robot secure package recognition using the encryption algorithm of Ridgelet transform with Rubik's cube principle and DenseNet-CNN based on the robot secure package recognition (RRCP-DenseNet). The performance of this proposed work is evaluated in the aspects of security and accuracy in the recognition of package images. The percentages of improvement from LeNet to DenseNet including raw input images were 16.5%, 7.26%, 21.41%, 23.12%, 9.65%, and 287.89% for raw input images and encrypted images using the Hill cipher, RSA, bit slicing, AES, RCP, and RRCP encryption algorithms, respectively. (c) 2022 SPIE and IS&T