Chrome Extension for Deepfake Detection, Deepfakes as a Wicked Problem 47 views

The rapid spread of artificial intelligence-generated media poses a major threat to digital authenticity. To address this challenge, my work investigates deepfakes through a dual lens involving the development of a real-time detection tool and a sociotechnical analysis of the moderation landscape. The technical capstone project focuses on building a Chrome extension that concurrently identifies manipulated media as a user browses the web. This system uses a client-server architecture where the frontend extracts video and picture frames directly from the browser and sent to a backend server. There, various preprocessing algorithms crop facial features and prepare the images for processing by a convolutional neural network. The model achieved 90% accuracy on the testing dataset, but live testing on YouTube yielded approximately 70% accuracy with a two-second latency. While the system has room for future improvements, this project demonstrates a feasible and practical solution for real-time detection of deepfakes. My STS research paper complements this technical work by framing deepfake moderation as a wicked problem. This framework describes social issues that lack definitive solutions because of conflicting stakeholder values and unintended consequences. The analysis demonstrates that while developers and content creators focus on innovation, platforms and policymakers must balance free expression against the risks of misinformation. Moreover, automated moderation can introduce problems like identity-based bias that harms underrepresented groups. These findings suggest that technical solutions alone are insufficient for resolving the general sociopolitical issues associated with synthetic media. Instead, the STS research paper advocates for hybrid governance combining detection technology with user education and transparent labeling. This system requires active involvement from all stakeholders: lawmakers, detection algorithm designers, social media companies, and users. By treating deepfake detection as an ongoing sociotechnical process, this work concludes that a sustainable defense requires both algorithmic improvement and enhanced digital literacy.

BS (Bachelor of Science)

deepfake; synthetic media; artificial intelligence; wicked problem

School of Engineering and Applied Science Bachelor of Science in Computer Science Technical Advisor: Rosanne Vrugtman STS Advisor: Travis Elliott

All rights reserved by the author (no additional license for public reuse)

2026-05-07