The papers in this special section focus on new solutions and research initiatives in the area of media authentication and forensics. Media manipulation is now a pressing societal problem with broad implications. In the past, it required significant skill to create compelling manipulations because editing tools, such as Adobe Photoshop, required experienced users to alter images convincingly. Over the last several years, the rise of machine learning-based technologies has dramatically lowered the skill necessary to create compelling manipulations. For example, Generative Adversarial Networks (GANs) can create photo-realistic faces with no skill by an end-user other than the ability to refresh a web page. Deepfake algorithms, such as autoencoders to swap faces in a video, can create manipulations much more easily than previous generation video tools. While these ML techniques have many positive uses, they have also been misused for darker purposes such as to perpetrate fraud, to create false personas, and to attack personal reputations. Clearly, these are pressing and important problems for which technical solutions must play a role.

中文翻译：

---

社论：媒体认证和取证——新的解决方案和研究机会


本专题部分的论文重点关注媒体身份验证和取证领域的新解决方案和研究计划。媒体操纵现在是一个具有广泛影响的紧迫社会问题。过去，创建令人信服的操作需要很高的技能，因为编辑工具（例如 Adob​​e Photoshop）需要经验丰富的用户才能令人信服地更改图像。在过去的几年里，基于机器学习的技术的兴起极大地降低了创建引人注目的操作所需的技能。例如，生成对抗网络（GAN）可以创建照片般逼真的面孔，最终用户除了刷新网页的能力之外不需要其他任何技能。 Deepfake 算法（例如在视频中交换面孔的自动编码器）可以比上一代视频工具更轻松地进行操作。虽然这些机器学习技术有许多积极的用途，但它们也被滥用于更黑暗的目的，例如实施欺诈、创建虚假人物角色以及攻击个人声誉。显然，这些都是紧迫而重要的问题，技术解决方案必须发挥作用。