Josef Kittler, former President of the International Association for Pattern Recognition, Fellow of the Royal Academy of Engineering, Distinguished Professor at the University of Surrey, IAPR Fellow, IEEE/IET Fellow

Presentation title:

Digital Content forensics in the context of large models

Speech abstract:

In the digital era, with the rapid development of artificial intelligence technology, especially the wide application of deep learning technology, the generation and editing of digital content has become more convenient and efficient. However, the double-edged nature of technology also brings new challenges in the field of digital content forensics. Generative large models, which can generate realistic text, images, audio and video, are likely to be widely used for malicious purposes such as false information and deep forgery, posing a threat to social order and information security. In the context of large models, forensics work becomes more complex and requires a higher level of technical means to cope with the continuous progress of counterfeiting technology. To address online disinformation and high quality fake content generated by large models, this report introduces several key technologies and a holistic approach to digital content forensics. This report focuses on the detection and forensics of traditional image tampering, the detection of portrait deep forgery, and the detection of the latest AIGC images and videos, as well as the detection and factual verification of disinformation that has spread widely on the Web. To generate the content for the large model, we also prospectively start from the source, edit the knowledge and limit the output content for the large model. These studies have been explored from the perspectives of generalization, interpretability, generating antagonistic game, etc., and have achieved remarkable results, providing important methods and ideas for guaranteeing the authenticity and credibility of digital content under the background of large models.

Xiong Hongkai, Distinguished Professor of Shanghai Jiao Tong University, Cheung Kong Scholar of the Ministry of Education, National Jieqing, leading talent of Ten thousand People Program, Deputy director of the "Visual Big Data" special Committee of the Chinese Society of Image and Graphics, and member of the Chinese Society of Electronics

Presentation title:

Digital Content forensics in the context of large models

Speech abstract:

In the digital era, with the rapid development of artificial intelligence technology, especially the wide application of deep learning technology, the generation and editing of digital content has become more convenient and efficient. However, the double-edged nature of technology also brings new challenges in the field of digital content forensics. Generative large models, which can generate realistic text, images, audio and video, are likely to be widely used for malicious purposes such as false information and deep forgery, posing a threat to social order and information security. In the context of large models, forensics work becomes more complex and requires a higher level of technical means to cope with the continuous progress of counterfeiting technology. To address online disinformation and high quality fake content generated by large models, this report introduces several key technologies and a holistic approach to digital content forensics. This report focuses on the detection and forensics of traditional image tampering, the detection of portrait deep forgery, and the detection of the latest AIGC images and videos, as well as the detection and factual verification of disinformation that has spread widely on the Web. To generate the content for the large model, we also prospectively start from the source, edit the knowledge and limit the output content for the large model. These studies have been explored from the perspectives of generalization, interpretability, generating antagonistic game, etc., and have achieved remarkable results, providing important methods and ideas for guaranteeing the authenticity and credibility of digital content under the background of large models.

Yang Jian, Professor of Nanjing University of Science and Technology, National Jieqing, Deputy director of Pattern Recognition Special Committee of Artificial Intelligence Society, director of Pattern Recognition Special Committee of Jiangsu Artificial Intelligence Society, IAPR Fellow, national leading talent

Presentation title:

Digital Content forensics in the context of large models

Speech abstract:

In the digital era, with the rapid development of artificial intelligence technology, especially the wide application of deep learning technology, the generation and editing of digital content has become more convenient and efficient. However, the double-edged nature of technology also brings new challenges in the field of digital content forensics. Generative large models, which can generate realistic text, images, audio and video, are likely to be widely used for malicious purposes such as false information and deep forgery, posing a threat to social order and information security. In the context of large models, forensics work becomes more complex and requires a higher level of technical means to cope with the continuous progress of counterfeiting technology. To address online disinformation and high quality fake content generated by large models, this report introduces several key technologies and a holistic approach to digital content forensics. This report focuses on the detection and forensics of traditional image tampering, the detection of portrait deep forgery, and the detection of the latest AIGC images and videos, as well as the detection and factual verification of disinformation that has spread widely on the Web. To generate the content for the large model, we also prospectively start from the source, edit the knowledge and limit the output content for the large model. These studies have been explored from the perspectives of generalization, interpretability, generating antagonistic game, etc., and have achieved remarkable results, providing important methods and ideas for guaranteeing the authenticity and credibility of digital content under the background of large models.

Chen Xilin, Director and Party Secretary of Institute of Computing Technology, Chinese Academy of Sciences, National Jie Qing, ACM/CCF/IAPR/IEEE Fellow, He has been the director of the Key Laboratory of Intelligent Information Processing and the Director of the International Cooperation Bureau of the Chinese Academy of Sciences

Presentation title:

Digital Content forensics in the context of large models

Speech abstract:

In the digital era, with the rapid development of artificial intelligence technology, especially the wide application of deep learning technology, the generation and editing of digital content has become more convenient and efficient. However, the double-edged nature of technology also brings new challenges in the field of digital content forensics. Generative large models, which can generate realistic text, images, audio and video, are likely to be widely used for malicious purposes such as false information and deep forgery, posing a threat to social order and information security. In the context of large models, forensics work becomes more complex and requires a higher level of technical means to cope with the continuous progress of counterfeiting technology. To address online disinformation and high quality fake content generated by large models, this report introduces several key technologies and a holistic approach to digital content forensics. This report focuses on the detection and forensics of traditional image tampering, the detection of portrait deep forgery, and the detection of the latest AIGC images and videos, as well as the detection and factual verification of disinformation that has spread widely on the Web. To generate the content for the large model, we also prospectively start from the source, edit the knowledge and limit the output content for the large model. These studies have been explored from the perspectives of generalization, interpretability, generating antagonistic game, etc., and have achieved remarkable results, providing important methods and ideas for guaranteeing the authenticity and credibility of digital content under the background of large models.