EAGER: Collaborative: Understanding How Manipulated Images Influence People

由于大量的硬件和软件工具大大降低了操纵数字图像所需的成本和工作量，恶意攻击者很容易通过计算机和社交网络发送篡改的图像，从而有目的地影响观众的意见、态度和行动，由此带来的风险和危险从未像现在这样严重。虽然越来越多的人意识到图像不再代表现实的真实证据，但在公众对视觉错误信息的脆弱性以及个人如何对图像真实性进行可信评估的研究方面存在空白。为了填补研究空白，本研究开展了一系列实证研究，以了解网络观众在形象可信度评价中使用的社会和认知启发式，以及这些评价如何影响他们的态度和行为。来自这些研究的数据将有助于预测观众最有可能接受网上图片被篡改的证据的方式，以及他们随后如何修正自己的情绪和信念。这项工作还着眼于网络攻击者可能利用社交媒体通过传播视觉错误信息来颠覆社会秩序的潜在方式，以及哪些策略可以有效地打击此类行为。这项工作的结果将为法医图像分析软件的设计提供信息，并将为新技术奠定基础，这些新技术将帮助互联网用户不断评估他们在网上收到的介导的视觉骗局和骗局的真实性。

As an abundance of hardware and software tools is dramatically decreasing the cost and effort required to manipulate digital images, the risks and dangers associated with malicious attackers easily routing doctored images through computer and social networks to purposefully influence viewers' opinions, attitudes, and actions have never been more severe. While there is a growing awareness that images no longer represent an authentic proof of reality, there is a gap in research about the public's vulnerabilities to visual misinformation and how individuals make credible evaluations about image authenticity. Filling the gap in research, this work conducts a series of empirical studies to find out the social and cognitive heuristics online viewers use in image credibility evaluation and how such evaluations influence their attitudes and behaviors. The data from these studies will help predict the ways by which viewers are most likely to accept evidence that online images have been manipulated and how they subsequently revise their emotions and beliefs surrounding them. This work also looks at potential ways cyber attackers could use social media to subvert social order by spreading visual misinformation, and what strategies would be effective in combating such behavior. Results from this work will inform the design of software for forensic image analysis and will lay the grounds for new technologies that help Internet users in continuously assessing the veracity of the mediated visual hoaxes and scams they receive online.