A Peguis student explored how reliable facial recognition technology really is in everyday situations as part of his project at the 2026 Canada-Wide Science Fair in Edmonton.
Keegan Anderson represented Manitoba at the national STEM competition with a project titled “The Effects of Lighting Conditions on the Accuracy and Security of Face ID,” examining how lighting and facial coverings affect the performance of smartphone facial recognition systems.
The 64th annual Canada-Wide Science Fair, organized by Youth Science Canada, took place May 23 to 30 at the Edmonton EXPO Centre and the University of Alberta. Nearly 400 finalists from across Canada competed for more than $2 million in scholarships, awards and prizes while showcasing innovative science, technology, engineering and mathematics projects.
Anderson’s project focused on a technology many people use daily to unlock their phones, but may not fully understand.
“Face ID is used every day to unlock phones, but how well does it work in different conditions?” Anderson explained in his project summary.
To find out, he tested Face ID performance in bright light, dim light and darkness while also experimenting with masks, sunglasses and regular glasses. He also tested the security of the system by attempting to unlock the phone using another person and a photograph of himself.
His findings showed Face ID performed best when the face was fully visible in bright lighting conditions.
“When important facial features are covered, such as with a mask or sunglasses, the accuracy decreases,” Anderson concluded.
The project found sunglasses had a greater impact on performance than regular prescription glasses because they block the eye area, while the combination of both a mask and sunglasses resulted in the lowest success rate because much of the face was hidden.
Although Face ID is designed to function in low light, Anderson found the system became less reliable in darker conditions.
He also discovered the technology appeared secure during his testing.
“Face ID did not unlock for another person or for a photo, which shows that it can tell the difference between the correct user and someone else,” he stated.
Anderson said the project highlights both the strengths and limitations of facial recognition technology in real-world situations and helps users better understand how dependable the systems are under different conditions.
If he continues the research in the future, Anderson said he would like to expand the testing to include more participants and different phone models to compare facial recognition systems across devices. He would also like to explore additional variables such as angles, distances and prosthetic disguises to further test how facial recognition technology responds in more complex real-world scenarios.
