Researchers show that your smartphone’s camera fingerprint allows anyone to track videos and pictures back to you

Posted on Nov 24, 2016 by Caleb Chen

Your smartphone has a unique camera fingerprint that can be used to match a photo or video with its recording device. A new app created by security researchers from Friedrich-Alexander-Universität Erlangen-Nuremberg and Politecnico Milano, CamFinger, allows users to learn more about camera fingerprinting by uploading a set of pictures to find out their own camera fingerprint. Private Internet Access spoke with Dominik Maier, one of the researchers at the computer security research group at Friedrich-Alexander-Universität Erlangen-Nuremberg. Dominik Maier believes that this camera fingerprint is a good candidate as a hardware token to secure banking transactions. He told PIA:

“I want to base a novel authentication scheme on top of this hardware token.”

While Dominik’s intentions may be pure, there are definitely malicious actors in this world that could be using camera fingerprinting technology for nefarious purposes. It’s long been known that each camera will have a unique pattern (fingerprint) known as a sensor pattern noise; sensor pattern noise research has been ongoing for about a decade. Since the camera fingerprint is created by checking each pixel, removing the EXIF header will not remove this piece of identifiable information from your photo. Additionally, cropping the image and using most basic Instagram or Snapchat filters will not get rid of the camera fingerprint. However, changing the resolution, which most social media sites do, or blurring and then deblurring the photo will scrub your picture of this camera fingerprint. However, as most videos are uploaded without resolution changes or blurring, they are particularly susceptible to revealing camera fingerprints.

Every camera has its own camera fingerprint

Even camera sensors made on the same factory line for the same phone manufacturer to be put in the same phone model will be slightly different from sensor to sensor. These slight variations between products are what cause the camera fingerprint. To find the fingerprint, each picture must be looked at pixel by pixel for patterns to emerge. Such techniques are already used in courtrooms today to match shooters of both guns and cameras to their shooting devices. More specifically, this works on every digital camera ever made – and similar techniques exist for analog cameras. There’s even printer stenography to track printed pages to the specific printer they came from. If you don’t think that these digital fingerprints will hold up in court, though, you’re due for a new thought. Digital camera fingerprints were first used as evidence in court in 2014 by Sussex police. When I asked Dominik whether governments were taking advantage of camera fingerprinting, he responded:

“[…]it would be far-fetched to believe governments don’t use it.”

What’s the worst case scenario with camera fingerprinting?

The disconcerting thought that Camfinger’s existence raises is that anyone, not just law enforcement, could know your camera fingerprint just by going through your camera. Does the government even need to source your camera fingerprint from the photos you do or don’t post online? It’s possible that companies will, of their own volition because of the incredible potential amounts of money involved, start storing the camera fingerprint of every phone they manufacture. What’s worse, planting your camera fingerprint onto an incriminating photograph is technically possible according to Dominik. While the field of digital image forensics is still evolving, the public’s understanding of what is possible will inevitably catch up.

Comments are closed.

4 Comments

  1. antimon555

    Strange, your reply shows up before I allow the site in NoScript, but after that it doesn’t. I guess implementing the noise should be done and/or checked by encryption experts – I’d imagine it would only take a tiny bit of predictability in the modification of the picture to still be able to identify the camera, due to the vast number of pixels…

    Also, I wonder about changing the resolution, wouldn’t that just cause a new unique pattern, that is the same for each camera as long as the source resolution and destination resolution is the same pair? Has that been tested?

    7 years ago
  2. Antimon555

    Would adding a low level of noise – slightly and randomly alter each pixel’s color and brightness – be enough to foil this technology?

    7 years ago
    1. Caleb Chen

      If done properly, possibly.

      7 years ago
    2. Lupo Čin

      “However, changing the resolution, which most social media sites do, or blurring and then deblurring the photo will scrub your picture of this camera fingerprint.” it’s literally written in the article

      7 years ago