Social Media Image Detection: Classification of Military Social Media Videos for Geographical Location Approximation
Loading...
Authors
Preti, Christian
Issue Date
2025-05-21
Type
Research Paper
Language
en_US
Keywords
Alternative Title
Abstract
The United States has seen growing threats in the information security domain. The distribution and analysis of Publicly Available Information (PAI) on social media networks (such as TikTok) has allowed adversaries to create profiles on service members. This data is “used for marketing” and “can be used to target an individual in their home, commit identity fraud, harassment, or target groups as part of an influence campaign” [4]. Additionally, the information is cheap to acquire; a study completed through Duke University (and funded by the United States Military Academy) found that “sensitive data about active-duty members of the military” was found “for as low as $0.12 per record” [12]. The advancements in Large Language Models (LLMs) and Vision-Language Models (VLMs) have created the opportunity for adversaries to compile data from social media platforms and determine approximate unit locations. This severely impacts Operational Security, mostly due to social media, which has “increased the speed and range of information, diffused power over information, and shifted socio-cultural norms” [7]. Overall, the need to understand this spread of information—and possibly use it against adversaries—has been realized across the globe and has become a genuine concern for nation states.