News authentication and tampered images: evaluating the photo-truth impact through image verification algorithms

• DOI: 10.1016/j.heliyon.2020.e05808
• Publication Date: 2020-12-24T22:44:23Z
• Authors: Anastasia N. Kats...; Antonios Gardikiotis; Nikolaos Tsipas; Charalampos A. Di...
• Keywords: Ground truth

Photos have been used as evident material in news reporting almost since the beginning of Journalism. In this context, manipulated or tampered pictures are very common part informing articles, today's misinformation crisis. The current paper investigates ability people to distinguish real from fake images. presented data derive two studies. Firstly, an online cross-sectional survey (N = 120) was conducted analyze ordinary human skills recognizing forgery attacks. target evaluate individuals' perception identifying visual content, therefore, investigate feasibility "crowdsourced validation". This last term refers process gathering fact-checking feedback multiple users, thus collaborating towards assembling pieces evidence on event. Secondly, given that contemporary veracity solutions coupled with both journalistic principles and technology developments, experiment phases employed: a) A repeated measures quantify associated abilities Media Image Experts 5 + 5) detecting tampering artifacts. latter case, image verification algorithms were put into core analysis procedure examine their impact authenticity assessment task. b) Apart conducting interview sessions selected experts proper guidance using tools, a second also deployed larger scale through 301), aiming at validating some initial findings. primary intent combined interpretation forensic services, offered real-world regarding comprehension utilization by people, involved everyday battle against misinformation. outcomes confirmed suspicion only few subjects had prior knowledge implicated algorithmic solutions. Although these assistive tools often lead controversial even contradictory conclusions, experimental treatment systematic training use boosted participants' performance. Overall, research findings indicate scores successful detections, relying exclusively observations, cannot be disregarded. Hence, ultimate challenge for "verification industry" should balance between automations experience, defending audience inaccurate information propagation.