Guobiao Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Guobiao Li has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Sheng Li
University of Virginia, Charlottesville, Virginia, USA
4 shared publications - Xinpeng Zhang
Fudan University, Shanghai, China
4 shared publications - Zhenxing Qian
Fudan University, Shanghai, China
4 shared publications Guangwei Tian
Department of Radiation Oncology, The First Affiliated Hospital of China Medical University, Shenyang, China
1 shared publication- Junqiang Huang
School of Computer Science, Fudan University, Shanghai, CN
1 shared publication - Nan Li
Department of Radiation Oncology, The First Affiliated Hospital of China Medical University, China
1 shared publication
Latest Publications
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 541-545
Toward Detecting Hidden Functionalities in Deep Learning Models
Deep functionality hiding is an emerging technique that embeds confidential or sensitive functions within seemingly benign deep learning models (DLMs), which perform ordinary machine learning tasks. This enables such models to execute covert tasks while remaining undetected. Despite the rapid progress in deep functionality hiding, countermeasures remain unexplored. In this paper, we propose Distribution Offset Analysis (DOA), a novel method for detecting hidden functionalities in DLMs. Our key insight is that the weight distribution of a benign DLM typically follows a Gaussian distribution, whereas a container DLM with hidden functionalities exhibits notable statistical deviations from this Gaussian pattern. In our methodology, we first compute the distributional distance (i.e., offsets) between the model's weights and an ideal Gaussian distribution. We then fuse these offsets with weight features into a unified representation, which is subsequently used to train a meta-classifier for hidden functionality detection. Through extensive experiments, we demonstrate the effectiveness of the proposed DOA method, which achieves an average detection rate of over 87% against existing state-of-the-art deep functionality hiding techniques.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 541-545
Toward Detecting Hidden Functionalities in Deep Learning Models
Deep functionality hiding is an emerging technique that embeds confidential or sensitive functions within seemingly benign deep learning models (DLMs), which perform ordinary machine learning tasks. This enables such models to execute covert tasks while remaining undetected. Despite the rapid progress in deep functionality hiding, countermeasures remain unexplored. In this paper, we propose Distribution Offset Analysis (DOA), a novel method for detecting hidden functionalities in DLMs. Our key insight is that the weight distribution of a benign DLM typically follows a Gaussian distribution, whereas a container DLM with hidden functionalities exhibits notable statistical deviations from this Gaussian pattern. In our methodology, we first compute the distributional distance (i.e., offsets) between the model's weights and an ideal Gaussian distribution. We then fuse these offsets with weight features into a unified representation, which is subsequently used to train a meta-classifier for hidden functionality detection. Through extensive experiments, we demonstrate the effectiveness of the proposed DOA method, which achieves an average detection rate of over 87% against existing state-of-the-art deep functionality hiding techniques.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (3) , 4559-4573
On Cover Independent Deep Neural Network Steganography
Steganography aims to hide secret data into a cover media by subtle perturbation. Recently, deep neural network (DNN) based steganography has attracted a lot of research interests. Most of the existing DNN-based steganographic schemes design the secret encoder and decoder in an asymmetrical manner, the performance of which is heavily dependent on the content of cover media. In addition, they are weak in preventing unauthorized data extraction from the stego-media (i.e., the media with hidden data). To address these two issues, we propose a novel DNN steganographic framework termed the Cover Independent DNN Steganography (CIDS). In our CIDS, we take advantage of the powerful generative models to obtain AI-generated cover media for both the sender and receiver according to a key. Then, we propose a pair of symmetrical secret encoder and decoder to conduct a bijective transformation between the secrets and perturbations. While the perturbation is combined with the cover media to produce the stego-media. Such a secret encoding/decoding strategy focuses only on the secrets and is independent of the cover media. We also propose to encrypt the perturbations to effectively prevent unauthorized data extraction. Comprehensive experiments demonstrate the advantage of our CIDS over the state-of-the-art approaches for image steganography.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (3) , 4559-4573
On Cover Independent Deep Neural Network Steganography
Steganography aims to hide secret data into a cover media by subtle perturbation. Recently, deep neural network (DNN) based steganography has attracted a lot of research interests. Most of the existing DNN-based steganographic schemes design the secret encoder and decoder in an asymmetrical manner, the performance of which is heavily dependent on the content of cover media. In addition, they are weak in preventing unauthorized data extraction from the stego-media (i.e., the media with hidden data). To address these two issues, we propose a novel DNN steganographic framework termed the Cover Independent DNN Steganography (CIDS). In our CIDS, we take advantage of the powerful generative models to obtain AI-generated cover media for both the sender and receiver according to a key. Then, we propose a pair of symmetrical secret encoder and decoder to conduct a bijective transformation between the secrets and perturbations. While the perturbation is combined with the cover media to produce the stego-media. Such a secret encoding/decoding strategy focuses only on the secrets and is independent of the cover media. We also propose to encrypt the perturbations to effectively prevent unauthorized data extraction. Comprehensive experiments demonstrate the advantage of our CIDS over the state-of-the-art approaches for image steganography.
- Open Access
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1
Emotion-Aware and Efficient Meme Sticker Dialogue Generation
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1
Open AccessEmotion-Aware and Efficient Meme Sticker Dialogue Generation
- Open Access
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 10238-10247
Cover-separable Fixed Neural Network Steganography via Deep Generative Models
Image steganography is the process of hiding secret data in a cover image by subtle perturbation. Recent studies show that it is feasible to use a fixed neural network for data embedding and extraction. Such Fixed Neural Network Steganography (FNNS) demonstrates favorable performance without the need for training networks, making it more practical for real-world applications. However, the stego-images generated by the existing FNNS methods exhibit high distortion, which is prone to be detected by steganalysis tools. To deal with this issue, we propose a Cover-separable Fixed Neural Network Steganography, namely Cs-FNNS. In Cs-FNNS, we propose a Steganographic Perturbation Search (SPS) algorithm to directly encode the secret data into an imperceptible perturbation, which is combined with an AI-generated cover image for transmission. Through accessing the same deep generative models, the receiver could reproduce the cover image using a pre-agreed key, to separate the perturbation in the stego-image for data decoding. such an encoding/decoding strategy focuses on the secret data and eliminates the disturbance of the cover images, hence achieving a better performance. We apply our Cs-FNNS to the steganographic field that hiding secret images within cover images. Through comprehensive experiments, we demonstrate the superior performance of the proposed method in terms of visual quality and undetectability. Moreover, we show the flexibility of our Cs-FNNS in terms of hiding multiple secret images for different receivers. Code is available at https://github.com/albblgb/Cs-FNNS
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 10238-10247
Open AccessCover-separable Fixed Neural Network Steganography via Deep Generative Models
Image steganography is the process of hiding secret data in a cover image by subtle perturbation. Recent studies show that it is feasible to use a fixed neural network for data embedding and extraction. Such Fixed Neural Network Steganography (FNNS) demonstrates favorable performance without the need for training networks, making it more practical for real-world applications. However, the stego-images generated by the existing FNNS methods exhibit high distortion, which is prone to be detected by steganalysis tools. To deal with this issue, we propose a Cover-separable Fixed Neural Network Steganography, namely Cs-FNNS. In Cs-FNNS, we propose a Steganographic Perturbation Search (SPS) algorithm to directly encode the secret data into an imperceptible perturbation, which is combined with an AI-generated cover image for transmission. Through accessing the same deep generative models, the receiver could reproduce the cover image using a pre-agreed key, to separate the perturbation in the stego-image for data decoding. such an encoding/decoding strategy focuses on the secret data and eliminates the disturbance of the cover images, hence achieving a better performance. We apply our Cs-FNNS to the steganographic field that hiding secret images within cover images. Through comprehensive experiments, we demonstrate the superior performance of the proposed method in terms of visual quality and undetectability. Moreover, we show the flexibility of our Cs-FNNS in terms of hiding multiple secret images for different receivers. Code is available at https://github.com/albblgb/Cs-FNNS
- Open Access
- 3 February 2022
- Published by Frontiers Media SA in Frontiers in Oncology
- Vol. 12, 797250
Comparison of Two Cyberknife Planning Approaches for Multiple Brain Metastases
Purpose: To compare the delivery efficiency, plan quality, and planned treatment volume (PTV) and normal brain dosimetry between different Cyberknife planning approaches for multiple brain metastases (MBM), and to evaluate the effects of the number of collimators on the related parameters.Methods: The study included 18 cases of MBM. The Cyberknife treatment plans were classified as Separate or Combined. For the Separate plan, each lesion was targeted by the collimator auto-selection method (Conformality 2/3 collimators). For the Combined plan, a PTV including all PTVs was targeted by the collimators. Monitor units (MUs), number of nodes and beams, estimated fraction treatment time (EFTT), new conformity index (nCI), dose gradient index (GI), homogeneity index (HI), PTV minimum/maximum dose (Dmax/Dmin), volume doses (D2% and D98%), maximum doses to lenses, optic nerves, and brainstem as well as normal brain 3, 6, 10, and 12 Gy (V3Gy–V12Gy) were compared.Results: Compared to the Combined plan, the Separate plan had fewer nodes and beams, shorter EFTT, smaller PTV Dmin, normal brain dose, and GI, and larger HI. The Separate plan with 2 collimators also had worse PTV coverage. In the Combined plan, more collimators increased beams, EFTT, GI, and normal brain dose but improved the PTV Dmin. Among treatments based on the Separate approach, there were obvious differences between plans for most of the items except the nCI. Fewer collimators resulted in significantly reduced beams, EFTT, PTV D98%, and normal brain dose with improved GI, although PTV Dmin and MUs were decreased while HI was increased.Conclusion: Both approaches met the requirements for SRS/HFSRT. We found that Separate plans improved treatment efficiency and normal tissue dosimetry.
- 3 February 2022
- Published by Frontiers Media SA in Frontiers in Oncology
- Vol. 12, 797250
Open AccessComparison of Two Cyberknife Planning Approaches for Multiple Brain Metastases
Purpose: To compare the delivery efficiency, plan quality, and planned treatment volume (PTV) and normal brain dosimetry between different Cyberknife planning approaches for multiple brain metastases (MBM), and to evaluate the effects of the number of collimators on the related parameters.Methods: The study included 18 cases of MBM. The Cyberknife treatment plans were classified as Separate or Combined. For the Separate plan, each lesion was targeted by the collimator auto-selection method (Conformality 2/3 collimators). For the Combined plan, a PTV including all PTVs was targeted by the collimators. Monitor units (MUs), number of nodes and beams, estimated fraction treatment time (EFTT), new conformity index (nCI), dose gradient index (GI), homogeneity index (HI), PTV minimum/maximum dose (Dmax/Dmin), volume doses (D2% and D98%), maximum doses to lenses, optic nerves, and brainstem as well as normal brain 3, 6, 10, and 12 Gy (V3Gy–V12Gy) were compared.Results: Compared to the Combined plan, the Separate plan had fewer nodes and beams, shorter EFTT, smaller PTV Dmin, normal brain dose, and GI, and larger HI. The Separate plan with 2 collimators also had worse PTV coverage. In the Combined plan, more collimators increased beams, EFTT, GI, and normal brain dose but improved the PTV Dmin. Among treatments based on the Separate approach, there were obvious differences between plans for most of the items except the nCI. Fewer collimators resulted in significantly reduced beams, EFTT, PTV D98%, and normal brain dose with improved GI, although PTV Dmin and MUs were decreased while HI was increased.Conclusion: Both approaches met the requirements for SRS/HFSRT. We found that Separate plans improved treatment efficiency and normal tissue dosimetry.
- 5 March 2020
- Published by SAGE Publications in DNA and Cell Biology
- Vol. 39 (3) , 417-427
Glycolysis-Based Genes Associated with the Clinical Outcome of Pancreatic Ductal Adenocarcinoma Identified by The Cancer Genome Atlas Data Analysis
Pancreatic ductal adenocarcinoma (PDAC) is one of the deadly tumors in digestive tract tumors. Although there has been advancement in PDAC treatment, its prognosis still remains unsatisfactory, mainly because of dismal diagnosis. This article aims to develop new prognostic factors related to energy metabolism in PDAC and to use these genes for novel risk stratification. Hundred fifty messenger RNA (mRNA) expression profiles and clinicopathological data of PDAC were downloaded from The Cancer Genome Atlas dataset. The glycolysis pathway was the significant pathway based on the gene set enrichment analysis. We chose the glycolysis pathway-related 176 genes for further analysis. Multivariate Cox regression analysis and forward stepwise Cox regression model established a novel three-gene glycolytic signature (including MET, B3GNT3, and SPAG4) for PDAC patients' prognosis prediction. All 150 patients were classified into two groups by the median risk score. High-risk group had a worse outcome compared to the low-risk group. The risk score was also significantly correlated with age and radiotherapy. A nomogram, including the glycolytic gene signature, has shown some clinical net benefit for overall survival prediction. We also validated the validity and reliability in the Puleo dataset. This novel gene expression signature may be involved in the pathophysiology and used for risk stratification and prognosis prediction in PDAC.
- 5 March 2020
- Published by SAGE Publications in DNA and Cell Biology
- Vol. 39 (3) , 417-427
Glycolysis-Based Genes Associated with the Clinical Outcome of Pancreatic Ductal Adenocarcinoma Identified by The Cancer Genome Atlas Data Analysis
Pancreatic ductal adenocarcinoma (PDAC) is one of the deadly tumors in digestive tract tumors. Although there has been advancement in PDAC treatment, its prognosis still remains unsatisfactory, mainly because of dismal diagnosis. This article aims to develop new prognostic factors related to energy metabolism in PDAC and to use these genes for novel risk stratification. Hundred fifty messenger RNA (mRNA) expression profiles and clinicopathological data of PDAC were downloaded from The Cancer Genome Atlas dataset. The glycolysis pathway was the significant pathway based on the gene set enrichment analysis. We chose the glycolysis pathway-related 176 genes for further analysis. Multivariate Cox regression analysis and forward stepwise Cox regression model established a novel three-gene glycolytic signature (including MET, B3GNT3, and SPAG4) for PDAC patients' prognosis prediction. All 150 patients were classified into two groups by the median risk score. High-risk group had a worse outcome compared to the low-risk group. The risk score was also significantly correlated with age and radiotherapy. A nomogram, including the glycolytic gene signature, has shown some clinical net benefit for overall survival prediction. We also validated the validity and reliability in the Puleo dataset. This novel gene expression signature may be involved in the pathophysiology and used for risk stratification and prognosis prediction in PDAC.