The study, led by Jakub Paplhám of the university’s Machine Learning Group, processed high-resolution imagery to determine if visual discrepancies in state-released media suggest the presence of a double, Novinky reported.
Researchers Compare Facial Recognition Models
The CTU team applied four distinct facial recognition models to a dataset of 142 photographs. This collection included both official imagery from the Korean Central News Agency (KCNA) and independent footage, as well as photos of Howard X, a well-known professional impersonator of the North Korean leader. By comparing these disparate sources, the researchers aimed to identify inconsistencies that traditional visual inspection might miss.
“All ten photographs from the years 2012 to 2026 are as similar to each other as photographs of one person usually are,” Paplhám told Novinky. However, he said that the software could not definitively rule out the use of a double. “If you try to distinguish Kim from his imitator, it always works. But if you don’t know the imitator exists, the model would assign him to Kim,” Paplhám explained.
Anthropologist Examines Permanent Facial Features
Forenzní antropoložka Petra Urbanová from Masaryk University approached the inquiry by analyzing stable biological markers rather than algorithmic probability. Her examination focused on permanent anatomical features, including a specific scar above the right eye, the alignment of the upper incisors, the gap between teeth, and the unique morphology of the ears.
“The same person is in all the pictures; those identifying marks are visible on almost all the photographs,” Urbanová told Novinky. She emphasized that replicating such granular, stable characteristics would be exceptionally difficult for any double. While factors like significant weight fluctuation, the natural aging process, or changes in facial expression can alter appearance, these underlying physical markers remain consistent across the analyzed timeline.
The 2021 Military Parade Imagery
Particular attention was directed toward images from the September 9, 2021, military parade, which previously fueled international speculation regarding the leader’s health and the potential use of a substitute. In these images, the subject appears noticeably slimmer and younger than in other public appearances.
Urbanová concluded that these visual differences were likely the result of digital manipulation rather than the presence of a different person. “It is not a double; it is an edited photograph where Kim Jong-un is depicted and characteristics that do not correspond to him have been added,” she said.
Her hypothesis suggests the original imagery may have been combined with photos of a thinner individual, though she said that a definitive forensic confirmation would require access to the original raw files and their metadata.
Skepticism Regarding Online Rumors
Koreanist Nina Špitálníková maintains that theories surrounding body doubles often originate from unverified claims on social media platforms. These narratives frequently gain traction during periods when the leader is absent from public view, such as the three-week period in 2020 that prompted widespread international rumors of his death.
“These rumors are mostly spread by trolls on the internet. And people who write about it don’t check where the original link came from,” Špitálníková explained to Novinky. She views such speculation as part of a recurring cycle of misinformation that lacks foundational evidence.
Unanswered Questions in Forensic Analysis
While the CTU team and independent experts found no evidence of a body double, the lack of access to the original, unedited digital files from North Korean state media remains a primary limitation. The forensic investigation continues to rely on publicly available, potentially altered imagery, meaning that the full extent of digital manipulation techniques used by state media remains unknown.
On September 9, 2021, the military parade occurred, marking the date that continues to serve as a focal point for researchers analyzing the leader’s visual history.
Related reading
