University of Groningen: Sensor defects are ideal for forensic camera analysis; Computer scientists have developed a new tool to connect digital media with its creators

In a project aimed at developing smart tools to combat child exploitation, University of Groningen Computer scientists have developed a system for analyzing noise from individual cameras. This information can be used to associate a video or photo with a specific camera. The results have been published in the journals SN Computer Science on . 4 June 2022 And the expert systems With running apps June 10, 2022.

Holland It is the leading distributor of digital content showing child sexual abuse, according to a report Internet Monitoring Corporation In 2019. To combat this type of abuse, forensic tools are needed to analyze digital content in order to identify photos or videos that contain suspicious content that is offensive to children. Another source of untapped information is noise in photos or video frames. as part of European Union project, University of Groningen Computer scientists, with colleagues from University of Lyon (Spain), a method for extracting and classifying noise from an image or video that reveals the “fingerprint” of the camera it was made with.

bullet

“You can compare it to the specific grooves on a fired bullet,” he says George AzopardiAssistant Professor in the Information Systems Research Group at Bernoulli Institute of MathematicsComputer science and artificial intelligence in University of Groningen. Each firearm produces a specific pattern on the bullet, so forensic experts can match a bullet found at a crime scene to a specific firearm, or link two bullets found at different crime scenes to the same weapon.

“Every camera has some flaws in the built-in sensors, which appear as image noise in all frames but are not visible to the naked eye,” Azzopardi explains. This results in camera noise. Guru Benabhaktola, PhD student at both Groningen and University of Lyon, developed a system for extracting and analyzing this noise. “In image recognition, classifiers are used to extract information about the shapes and textures of objects in the image to identify a scene,” Bennabhaktula says. “We used these classifiers to extract camera-specific noise.”

law enforcement

He created a computational model to extract camera noise from video frames captured with 28 different cameras, taken from the publicly available VISION dataset, and used it to train a convolutional neural network. Next, test if the trained system can recognize tires made by the same camera. “It turns out we can do this with an accuracy of 72 percent,” Bennabhaktula says. It also points out that noise can be unique to a brand of camera, a specific type, and individual cameras. In another set of experiments, he achieved 99 percent accuracy in classifying 18 camera models using images from the publicly available Dresden dataset.

His work is part of European Union The 4NSEEK project, in which scientists and law enforcement agencies collaborated to develop smart tools to help combat child exploitation. Azzopardi: “Each group was responsible for developing a specific forensic tool.” The model created by Bennabhaktula could have such a practical use. If police find a camera of a suspected child abuser, they can link it to photos or videos on storage devices.

Challenges

Bennabhaktula adds that the model is scalable. Using only five random frames from a video, it is possible to rate five videos per second. The classifier used in the model has been used by others to distinguish between more than 10,000 different classes of other computer vision applications. This means that the classifier can compare the noise from tens of thousands of cameras. The 4NSEEK project is now over, but Azzopardi remains in contact with forensic professionals and law enforcement agencies to continue this line of research. We are also working to determine the similarity of the source between a pair of images, which faces different challenges. This will form our next paper on the topic.

References:

Guru Swaroop Benabhaktola, Derek TimmermanAnd the Enrique Allegri And the George Azopardi: Select the source camera device from the video clips. SN Computer Science, 4 June 2022.

Guru Swaroop Benabhaktola, Enrique AllegriDimka Karastoyanova and George Azopardi: Determination of the camera model based on forensic traces extracted from homogeneous patches. expert systems with apps, June 10, 2022.

Leave a Reply

%d bloggers like this: