The Impact of Multi-Camera Smart Phones on Source Camera Identification

An investigator has a questioned image from an unknown source and wants to determine whether it came from a camera on a person of interest’s smartphone. This scenario is referred to as source camera identification. Researchers discovered that slight imperfections in the photosites of a camera’s sensor array act as an identifying feature or camera fingerprint in this scenario. They showed that in most circumstances, these imperfections in the sensor array carry through the image-capturing process and remain in the noise level of the image’s pixel values, where it is unnoticeable to the human eye but detectable through computer processing. Source camera identification methods capture reference images with the person of interest’s camera, extract the noise residuals from each reference image, and average the noise residuals to estimate the camera fingerprint. The noise residual of the questioned image is used to estimate the unknown source camera. Then the two camera fingerprints are compared. With only a few exceptions, previous work in source camera identification relied on image datasets comprised solely of images from DSLR cameras or the main rear cameras of mobile phones. Nowadays, most major smartphone brands offer a front-facing selfie camera and two or more rear cameras. The Center for Statistics and Applications in Forensic Evidence (CSAFE) recently purchased ten iPhone 11 Pros, ten iPhone 12 Pros, ten iPhone 14 Pros, ten Samsung Galaxy Note 10s, ten Samsung Galaxy S20s, and ten Samsung Galaxy S21s. These sixty phones have a front selfie camera, a telephoto camera, a wide-angle camera, and an ultra-wide-angle camera. CSAFE collected one hundred flatfield and one hundred natural scene images from each camera on each phone to build a large, authenticated image database. This database will be made available for free to the forensic science community. We used this image database to develop and explore source camera identification methods on multi-camera phones. This presentation will focus on our insights from this research.