Partial facial recognition study

An OSINT investigation by Yoni and Alexis Martins explored how partial facial recognition tools can identify individuals even when parts of their face are hidden. They used a promotional video from the Ministry of Defense featuring a soldier named Clarisse, whose face was partially covered by a scarf but whose eyes, eyebrows, forehead, and hairline were visible.

High identification scores achieved

The researchers compared Clarisse’s partially concealed image with a blurred full-face photo of the same person. Using an internal tool, they achieved a similarity score of 90 percent with the scarf in place and 97 percent when the lower face was excluded from both images. These results suggest that partial facial coverage does not necessarily prevent accurate identification.

Limitations and operational security

The study highlights that identification scores depend on the specific tools used and are not universally standardized. It also stresses the importance of operational security (OPSEC) to minimize the risk of being identified. Measures like wearing sunglasses can complicate eye analysis but do not ensure complete anonymity.

Implications for public information handling

The researchers caution that photographs and videos should be carefully handled before publication. The accumulation of public information can lead to identification through cross-referencing, which poses risks for both individuals and organizations. This applies to military personnel, police officers, gendarmes, and other specialized units.

Facial recognition system factors

The effectiveness of facial recognition systems depends on image quality, resolution, lighting, and the visible characteristics of the person. Tools like Toolpie allow for grouping details, and artificial intelligence has improved the ability to analyze and organize information. This can be used by both legitimate investigators and potential malicious actors.