Juan Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Juan Zhang has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
Jiaen Hu
Beijing Institute of Graphic Communication
18 shared publicationsZichen Li
Beijing Institute of Graphic Communication
18 shared publications- Qinghao Wu
School of Information Engineering, Beijing Institute of Graphic Communication, Beijing 102600, China
1 shared publication - Zichen Li
School of Information Engineering, Beijing Institute of Graphic Communication, Beijing 102600, China
1 shared publication
Latest Publications
- Open Access
- 30 July 2025
- Published by Springer Nature in Research Square
Breaking Barriers: Multi-Agent Prompt Fusion for Automated LLM Jailbreaks
With the widespread deployment of Large Language Models (LLMs) across various natural language processing tasks, their potential security vulnerabilities have become increasingly prominent, emerging as a crucial issue in AI safety research. Although various jailbreak attack methods have been proposed to expose risks related to generation control and content safety, current studies mostly focus on isolated attack strategies, lacking multi-agent coordination mechanisms and thus failing to comprehensively probe the behavioral boundaries of models. To systematically explore the multidimensional risks of LLMs under jailbreak defense scenarios,this paper proposes a collaborative attack framework based on energy function modeling and multiagent game mechanisms. The framework treats suffix generation, input reconstruction, and context reshaping attacks as three independent agents, each capturing target constraints from distinct attack perspectives. To facilitate effective collaboration among these strategies, a game-theoretic dynamic optimization process is used, enabling agents to evolve jointly and adaptively adjust their respective contributions based on performance feedback. To enhance the linguistic naturalness and attack potency of the generated texts,we introduce a semantic fusion module assisted by a language model,optimizing semantic coherence and expression fluency while preserving adversarial goals. Furthermore,a multidimensional reward function system based on attack success rate (ASR),language perplexity (PPL),and semantic similarity is designed to continuously optimize the fusion strategies within a reinforcement learning framework. Experimental results demonstrate that our method significantly improves attack success rates and text naturalness over multiple attack rounds,producing adversarial samples with stronger stealth and usability. Our findings effectively reveal the current blind spots in LLM defenses under multi-dimensional inputs, advancing research in LLM security and adversarial defense.
- 30 July 2025
- Published by Springer Nature in Research Square
Open AccessBreaking Barriers: Multi-Agent Prompt Fusion for Automated LLM Jailbreaks
With the widespread deployment of Large Language Models (LLMs) across various natural language processing tasks, their potential security vulnerabilities have become increasingly prominent, emerging as a crucial issue in AI safety research. Although various jailbreak attack methods have been proposed to expose risks related to generation control and content safety, current studies mostly focus on isolated attack strategies, lacking multi-agent coordination mechanisms and thus failing to comprehensively probe the behavioral boundaries of models. To systematically explore the multidimensional risks of LLMs under jailbreak defense scenarios,this paper proposes a collaborative attack framework based on energy function modeling and multiagent game mechanisms. The framework treats suffix generation, input reconstruction, and context reshaping attacks as three independent agents, each capturing target constraints from distinct attack perspectives. To facilitate effective collaboration among these strategies, a game-theoretic dynamic optimization process is used, enabling agents to evolve jointly and adaptively adjust their respective contributions based on performance feedback. To enhance the linguistic naturalness and attack potency of the generated texts,we introduce a semantic fusion module assisted by a language model,optimizing semantic coherence and expression fluency while preserving adversarial goals. Furthermore,a multidimensional reward function system based on attack success rate (ASR),language perplexity (PPL),and semantic similarity is designed to continuously optimize the fusion strategies within a reinforcement learning framework. Experimental results demonstrate that our method significantly improves attack success rates and text naturalness over multiple attack rounds,producing adversarial samples with stronger stealth and usability. Our findings effectively reveal the current blind spots in LLM defenses under multi-dimensional inputs, advancing research in LLM security and adversarial defense.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Table 1: Statistical data of the dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessTable 1: Statistical data of the dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Table 3: Model performance improvement table.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessTable 3: Model performance improvement table.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 8: The effect of varying batch size on the model’s training performance.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 8: The effect of varying batch size on the model’s training performance.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Table 2: Performance of different methods on PHEME and GossipCop datasets.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessTable 2: Performance of different methods on PHEME and GossipCop datasets.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 10: Comparison of model performance on the GossipCop dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 10: Comparison of model performance on the GossipCop dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 7: The effect of varying learning rates on the model’s training performance.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 7: The effect of varying learning rates on the model’s training performance.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 6: Temporal sequence penetration fusion module.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 6: Temporal sequence penetration fusion module.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 5: Temporal sequence penetration fusion module.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 5: Temporal sequence penetration fusion module.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- Open Access
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Figure 9: Comparison of model performance on the PHEME dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.
- 7 July 2025
- Published by PeerJ
- Vol. 11, e2998
Open AccessFigure 9: Comparison of model performance on the PHEME dataset.
As the internet continues to evolve rapidly and social media becomes increasingly prevalent, the ways people access information has become increasingly diverse. However, the proliferation of fake news has emerged as a critical problem, presenting major challenges to the integrity of the information ecosystem. To address the complex propagation mechanisms of fake news, existing studies leverage multi-modal information and dynamic propagation social graphs for effective detection. Nonetheless, capturing the temporal relationships of propagation nodes in dynamic social networks accurately and dynamically integrating multi-modal information for improved detection accuracy remains a technical challenge. In response, This study proposes a multimodal approach to fake news detection—the dynamic temporal network (DTN) model. Firstly, this model designs a time similarity strength metric to measure the temporal similarity among nodes in propagation sequences and introduces a weighting mechanism to dynamically fuse multi-modal information. Secondly, it constructs a social propagation graph model, enhancing node representation through the dynamic variations of time similarity and graph structure, and utilizes the Transformer encoder to extract the overall semantic features of news propagation. Furthermore, the model views the news propagation process as a complex system, analyzing the temporal dynamics of news in real social networks, effectively revealing the abnormal propagation patterns of fake news. Further analysis demonstrates that the proposed DTN model exhibits high accuracy and effectiveness in multi-modal fake news detection.