23 January 2020 in IEEE Transactions on Industrial Informatics
We propose the PSDRNN and TriPSDRNN schemes that employ the explicit feature extraction before the Deep Recurrent Neural Network (DRNN) model. Specifically, considering that Power Spectral Density (PSD) feature can capture the frequency characteristics and meanwhile retain the successive time characteristics of data gathered from smartphone accelerometer, PSD feature vectors are respectively extracted from linear accelerations and tri-axle accelerations, and explicitly used as the input data to the following DRNN classification model. Thorough experiments based on a real dataset demonstrate that PSDRNN can achieve the comparable effectiveness as the xyz-DRNN (the best accurate DRNN based HAR scheme using only acceleration data), and the average recognition and training time were reduced by 56% and 80% respectively. Moreover, TriPSDRNN advantages over the xyz-DRNN in terms of recognition accuracy, and the running time is lower than xyz-DRNN. Besides, PSDRNN scheme can achieve superiority in the recognition of complex transition activities.
8 October 2024 in IEEE Transactions on Industrial Informatics
Computer vision (CV) applications empower various Internet of Things (IoT) scenarios. However, their advancements in image generation and manipulation tools make it increasingly easy to produce highly deceptive forged images, escalating the risk of image forgery. Cryptography-based methods can secure images but cannot support direct CV applications with compromised visual legibility. Existing generative adversarial network (GAN)-based steganography methods can effectively facilitate CV applications and image forgery prevention with high indistinguishability between stego and cover images. However, they are inefficient in resource-constrained IoT scenarios. Therefore, we propose a lightweight image forgery prevention scheme for IoT using GAN-based steganography. Our scheme embeds identity data within images. If forged, it fails to recover, triggering alerts. Our scheme can significantly improve efficiency with a lightweight generator designed by incorporating blueprint separable convolutions, sum connections and discrete wavelet transform while ensuring high effectiveness. Real-world IoT experimental results demonstrate this.
25 February 2025 in IEEE Transactions on Circuits and Systems for Video Technology
Artificial Intelligence Generated Content (AIGC) has created a fertile ground for image steganography. Existing Coverless Image Steganography (CIS) methods rely on image semantics to encode secrets, transmitting stego images without embedding, inherently resisting steganalysis. However, constructing CIS Datasets (CISDs) for these methods demands excessive resources, making them impractical for communication. Moreover, achieving low cost and high security is unattainable under these conditions. Therefore, we propose a CIS method based on semantic-controlled text-to-image generation. Our method disguises users as typical AIGC community members utilizing mainstream black-box text-to-image generation with Stable Diffusion (SD). During pre-processing, plain prompts, derived from dialogues with a large language model, are divided into coded and uncoded prompts through our encryption process, where a secret key determines coded prompts. In communication, confusion prompts are selected from uncoded and coded prompts, excluding those determined by secrets. Subsequently, our stego shuffling process combines topic, secret, and confusion prompts to produce stego prompt sets. Diverse stego images maintaining visual topic consistency are generated from these sets using SD with generation seeds indicating transmission order. By introducing confusion prompts, our method is secure from recognition when revealing stego prompts. Experimental results demonstrate our method achieves low communication costs and enhances communication security.
10 October 2026 in Multimedia Systems
19 September 2026 in Signal Processing