Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/238060 
Authors: 
Year of Publication: 
2021
Series/Report no.: 
23rd Biennial Conference of the International Telecommunications Society (ITS): "Digital societies and industrial transformations: Policies, markets, and technologies in a post-Covid world", Online Conference / Gothenburg, Sweden, 21st-23rd June, 2021
Publisher: 
International Telecommunications Society (ITS), Calgary
Abstract: 
The development of artificial intelligence (AI) technology has made it easy for users to generate hyper-realistic fake media content, and its most representative by-product is called deepfake. However, considerable attention has been paid to the adverse effects of deepfakes as they are tightly connected to the production of fake news, financial frauds, or fake pornographies. The misuse of deepfakes led to a series of deepfake prevention studies, but most were post-detection methods. This study thus investigated deepfake malfunction-inducing technology that may forestall the generation of deepfake with PGD attack. In the next part of the study, overall preferences and intention to use were measured as people's responses to this technology. An online survey especially targeting those exposed to various media like social media influencers (SMIs), was conducted. The deepfakes started to malfunction after adding 0.009 levels of an adversarial noise as a preventive mechanism. From a technical viewpoint, higher noise was a more effective way to prevent deepfake synthesis, but from the user's viewpoint, noise as high as 0.03 was found to be appropriate. Individuals' intention to use was tested with Bulgurcu's ISP compliance model. It was found that SMIs' predictive evaluations on the cost and benefit of this technology influence their attitude, and consequently, their intention to use it. This study shows the value of collaborative studies of AI-based privacy security domain and media industry domain. It also expands the scope of the framework with thorough hypothetical testing in the deepfake context.
Subjects: 
Deepfake
Adversarial noise
Image quality
Intention to use
Document Type: 
Conference Paper

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.