基于多模态信息融合的深度伪造检测

Deepfake Detection Based on Multi-mode Information Fusion

深度伪造技术的快速发展和应用给国家和社会安全、个人信息数据、企业安全等造成了潜在威胁。从电网企业的实际业务场景出发,针对现有深度伪造检测方法的泛化能力不足问题,文章提出一种基于多模态信息融合的深度伪造检测方法,为保护电网企业合法权益提供支撑。提出的方法在基准人脸伪造数据集FaceForensics++(FF++)进行验证评估,结果表明该方法在人脸伪造检测方面性能优良、鲁棒性好,同时多模态信息融合使得模型泛化性得到了很好的提升。

The rapid development and application of deep forgery generation technology have posed potential threats to personal privacy data,social stability,national and enterprise security,etc.Starting from the actual business scenarios of power grid enterprises,and aiming at the insufficient generalization ability of existing deep forgery detection methods,this paper proposes a deep forgery detection method based on multimodal information fusion to provide support for protecting the legitimate rights and interests of power grid enterprises.The method proposed in this paper is verified and evaluated in the benchmark face forgery dataset FaceForensics++(FF++).The experimental results show that the method has good performance and robustness in face forgery detection,and multimodal information fusion improves the generalization of the model.