基于生成对抗网络和对比学习的假新闻检测方法研究
录用日期: 2024-06-03
网络出版日期: 2024-11-16
基金资助
国家自然科学基金(62172303)
版权
An investigation into fake news detection methodology based on generative adversarial networks and contrastive learning
Accepted date: 2024-06-03
Online published: 2024-11-16
Copyright
社交媒体作为信息获取的主要途径,其假新闻问题日益严重。假新闻检测任务的重要挑战之一是确保模型能够及时响应新出现的事件,并在有限时间内完成检测任务,这要求模型具备高效的实时性和对新事件的快速适应能力,与此同时,多模态假新闻检测技术作为未来的重要发展方向也值得关注。针对上述挑战,提出了一种多模态假新闻检测模型ADSCL,利用卷积神经网络提取文本和图像的语义特征,并通过多层联合注意力机制进行融合。针对新事件的及时响应需求,引入生成对抗网络和对比学习,从大量数据中提取可转移特征,提高泛化能力。同时,通过对抗性训练增强模型鲁棒性。实验结果表明,ADSCL模型有效提升了假新闻检测能力,验证了多模态融合和对抗性方法在新闻检测任务上的优越性。
吴聪 , 孟敏智 , 郑炜 , 何琨 , 纪守领 . 基于生成对抗网络和对比学习的假新闻检测方法研究[J]. 网络空间安全科学学报, 2024 , 2(3) : 27 -40 . DOI: 10.20172/j.issn.2097-3136.240303
As the primary avenue for information retrieval, social media increasingly grapples with the issue of fake news. A key challenge in fake news detection is ensuring models can promptly respond to emerging events and complete detection tasks within a limited timeframe. This necessitates models with efficient real-time capabilities and rapid adaptation to new events. Meanwhile, the development of multimodal fake news detection technology as a crucial future direction also merits attention. Addressing these challenges, a multimodal fake news detection model, ADSCL was proposed. Leveraging convolutional neural networks, the model extracts semantic features from text and images, integrating them through a multi-layered joint attention mechanism. To meet the demand for timely responses to new events, the model introduces generative adversarial networks and contrastive learning to extract transferable features from extensive data, enhancing generalization capability. Furthermore, adversarial training enhances model robustness. Experimental results demonstrate that the ADSCL model effectively enhances the capability of fake news detection, validating the superiority of multimodal fusion and adversarial methods in news verification tasks.
表 1 数据集统计结果Table 1 Statistical Results of the Dataset |
| 统计数据 | 推特/条 | 微博/条 | |
| 训练集 | 谣言 | ||
| 非谣言 | |||
| 测试集 | 谣言 | 884 | |
| 非谣言 | 456 | 524 | |
| 总计 | |||
表 2 总体性能实验结果Table 2 Overall Performance Experiment Results |
| 数据集 | 模型 | 准确率 | 精确率 | 召回率 | F1 Score |
| 微博数据集 | Text Visual | 0.691 0.638 | 0.569 0.510 | 0.697 0.672 | 0.627 0.580 |
| VQA SAFE | 0.727 0.717 | 0.594 0.591 | 0.845 0.773 | 0.698 0.670 | |
| att-RNN | 0.779 | 0.663 | 0.826 | 0.736 | |
| ADSCL- | 0.800 | 0.658 | 0.850 | 0.742 | |
| ADSCL | 0.853 | 0.731 | 0.892 | 0.804 | |
| 推特数据集 | Text Visual | 0.547 0.598 | 0.338 0.390 | 0.502 0.566 | 0.404 0.462 |
| VQA SAFE | 0.662 0.646 | 0.467 0.451 | 0.750 0.735 | 0.576 0.559 | |
| att-RNN | 0.709 | 0.515 | 0.809 | 0.627 | |
| ADSCL- | 0.738 | 0.547 | 0.822 | 0.657 | |
| ADSCL | 0.759 | 0.571 | 0.844 | 0.681 |
表 3 生成对抗网络消融实验结果Table 3 Ablation experiment results for GANs |
| 数据集 | 模型 | 准确率 | 精确率 | 召回率 | F1-Score |
| 实验训练集 | Text | 52.2/% | 34.6/% | 46.5/% | 39.7 |
| Text+ | 57.2/% | 40.1/% | 54.3/% | 46.1 | |
| Visual | 50.7/% | 33.2/% | 45.5/% | 38.4 | |
| Visual+ | 55.5/% | 38.4/% | 52.2/% | 44.3 | |
| att-RNN | 75.9/% | 60.3/% | 84.0/% | 70.2 | |
| ADSCL- | 75.1/% | 60.3/% | 76.9/% | 67.6 | |
| ADSCL | 82.1/% | 70.5/% | 89.3/% | 78.8 |
表 4 对抗样本消融实验结果Table 4 Ablation Results for Adversarial Examples |
| 数据集 | 模型 | 准确率 | 精确率 | 召回率 | |
| 对抗测试集 | Text Visual | 55.3/% 54.3/% | 37.6/% 36.8/% | 48.9/% 47.4/% | 42.5 41.2 |
| VQA SAFE | 65.4/% 63.1/% | 49.1/% 46.3/% | 60.1/% 56.5/% | 54.0 50.9 | |
| att-RNN | 75.2/% | 60.1/% | 79.1/% | 68.3 | |
| ADSCL* | 46.3/% | 30.5/% | 46.2/% | 36.8 | |
| ADSCL | 84.2/% | 71.6/% | 88.6/% | 79.2 |
| 1 |
MA J, GAO W, WONG K F. Rumor detection on twitter with tree-structured recursive neural networks[C]//Association for Computational Linguistics, 2018.
|
| 2 |
LIN H,ZHANG X,FU X. A graph convolutional encoder and decoder model for rumor detection[C]//2020 IEEE 7th International Conference on Data Science and Advanced Analytics (DSAA),2020 :300-306.
|
| 3 |
BIAN T,XIAO X,XU T,et al. Rumor detection on social media with bi-directional graph convolutional networks[C]//Proceedings of the AAAI conference on artificial intelligence ,2020 :549-556.
|
| 4 |
韩雪明,贾彩燕,李轩涯,等.传播树结构结点及路径双注意力谣言检测模型[J].计算机科学,2023,50(4):22-31.
HAN X M,JIA C Y,LI X Y,et al. Dual-attention network model on propagation tree structures for rumor detection[J]. Computer Science,223,50(4): 22-31.
|
| 5 |
MA J , GAO W , MITRA P. Detecting rumors from microblogswith recurrent neural networks[C]//Prceeding of the 25th International Joint Conference on Artificial Intelligence,2016: 3818-3824.
|
| 6 |
KHOO L M S,CHIEU H L,QIAN Z,et al. Interpretable rumor detection in microblogs by attending to user interactions[C]//Proceeding of the AAAI Conference on Artificial Intelligence,AAAI 2020,2020,34(5) :8783-8790.
|
| 7 |
YU F,LIU Q,WU S,et al. A convolutional approach for misinformation identification[C]// Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence,2017 :3901-3907.
|
| 8 |
LIU Y,WU Y B. Early detection of fake news on social media through propagation path classification with recurrent and convolutional networks[C]// Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence,2018 :354-361.
|
| 9 |
GUO H,CAO J,ZHANG Y,et al. Rumor detection with hierarchical social attention network[C]// Proceedings of the 27th ACM international conference on information and knowledge management,2018 :943-951.
|
| 10 |
LIN X,LIAO X,XU T,et al. Rumor detection with hierarchical recurrent convolutional neural network[C]// Natural Language Processing and Chinese Computing:8th CCF International Conference,NLPCC 2019:338-348.
|
| 11 |
DONG S,QIAN Z,LI P,et al. Rumor detection on hierarchical attention network with user and sentiment information[C]// Natural Language Processing and Chinese Computing:9th CCF International Conference,2020 :366-377.
|
| 12 |
MA J,GAO W,WONG K F. Detect rumors on twitter by promoting information campaigns with generative adversarial learning[C]// The World Wide Web Conference,2019 :3049-3055.
|
| 13 |
Nikam S S, Dalvi R. Machine learning algorithm based model for classification of fake news on twitter[C]//2020 Fourth International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC). IEEE,2020: 1-4.
|
| 14 |
PRZYBYLA P. Capturing the style of fake news[C]// The Thirty-Fourth AAAI Conference on Artificial Intelligence,2020 :490-497.
|
| 15 |
SHU K, SLIVA A, WANG S, et al. Fake news detection on social media: A data mining perspective[J]. ACM SIGKDD Explorations Newsletter, 2017, 19 (1): 22- 36.
|
| 16 |
JIN Z, CAO J, ZHANG Y, et al. Novel visual and statistical image features for microblogs news verification[J]. IEEE Transactions on Multimedia, 2016, 19 (3): 598- 608.
|
| 17 |
JIN Z,CAO J,GUO H,et al. Multimodal fusion with recurrent neural networks for rumor detection on microblogs[C]// Proceedings of the 25th ACM international conference on Multimedia,2017 :795-816.
|
| 18 |
QI P,CAO J,YANG T,et al. Exploiting multi-domain visual information for fake news detection[C]// 2019 IEEE International Conference on Data Mining (ICDM),2019 :518-527.
|
| 19 |
ZHANG T,WANG D,CHEN H,et al. BDANN:BERT-based domain adaptation neural network for multi-modal fake news detection[C]// 2020 International Joint Conference on Neural Networks (IJCNN),2020 :1-8.
|
| 20 |
DEVLIN J,CHANG M-W,LEE K,et al. BERT:Pre-training of deep bidirectional transformers for language understanding[C]// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies,2019.
|
| 21 |
ZHOU Y, YANG Y, YING Q, et al. Multi-modal fake news detection on social media via multi-grained information fusion[J]. arXiv preprint arXiv:, 2304, 00827, 2023.
|
| 22 |
RADFORD A,KIM J W,HALLACY C,et al. Learning transferable visual models from natural language supervision[C]//International Conference on Machine Learning,2021 :8748-8763.
|
| 23 |
CHEN Y. Convolutional neural network for sentence classification[D]. University of Waterloo,2015.
|
| 24 |
SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:, 1409, 1556, 2014.
|
| 25 |
MADRY A, MAKELOV A, SCHMIDT L, et al. Towards deep learning models resistant to adversarial attacks[J]. arXiv preprint arXiv:, 1706, 06083, 2017.
|
| 26 |
ZHOU X,WU J,ZAFARANI R. Similarity-aware multi-modal fake news detection[C]//Pacific-Asia Conference on knowledge discovery and data mining. Cham:Springer International Publishing,2020:354-367.
|
| 27 |
JIN Z,CAO J,GUO H,et al. Multimodal fusion with recurrent neural networks for rumor detection on microblogs[C]// MM’ 17 :Proceedings of the 25th ACM International Conference on Multimedia,2017:795–816.
|
/
| 〈 |
|
〉 |