Source Camera Identification: a Distributed Computing Approach Using Hadoop

This article is all about the distributed computing approach to identify the source of images obtained from digital cameras by some methods like Apache Hadoop. The creation of such methods is an urgent task for forensic science and for Managed LAN Network/Managed Network Services that are liable sometimes to assist law enforcement agencies as their requirement, for example, in matters of copyright protection. The proposed approach is based on the search and comparison of the spatial location of particular pixels (defects in a photosensitive sensor) that are unique for each digital camera. Under the special pixels in the article refers to the so-called “broken” and “hot” pixels. Experiments have shown the potential use of the proposed method in identifying the source of digital images.

Due to the rapid development of methods for distributing digital materials, information about their authorship is often lost. Therefore, it becomes relevant to develop methods that establish the authenticity of digital photographs and cameras. The analysis showed the presence of only single implementations of such methods. The proposed authentication method is based on the analysis of the unique location of defective pixels for each instance of a digital camera.

In the matrix (a set of photosensitive sensors) of digital cameras, there are a certain number of specific pixels on the matrix. Special pixels in the article will be called “broken” and “hot” pixels.

By “broken” pixels (the official name is defective pixels) means a defect in an electronic device that perceives or reproduces an image and has a pixel structure. Manifested in the immutability of the output signal of several pixels.

Hot pixels - a defect in which the value of the output signal has an incorrect dependence on the input, or the output signal is most dependent on other factors (temperature, the value of neighboring pixels).

“Dependent” pixels are a slang term meaning a specific pixel is dependent on the value of its surrounding. As a rule, such pixels are invisible on realistic images. The cause of such a defect is a malfunction of the matrix element, in the case of a “broken” pixel - a constant malfunction.

In the case of a matrix of a digital camera, camcorder, scanner, document camera, or other image pickup device, the output signal value is the numeric value of the corresponding pixel in the digital camera file. In the overwhelming majority of digital cameras, masking of “dead” pixels is provided by interpolating their values from neighboring ones (thereby turning them into “dependent”).

In the physical aspects of the occurrence of defective pixels on photosensitive arrays of various types are considered. At the same time, the paper does not give constructive suggestions for developing methods for searching for defective pixels and methods for comparing them from received images.

In contrast to the methods for verification of the sensor - the source camera, where a set of images is required to build a template, in the proposed approach, the map of defective pixels is built on one presented image.

There are various ways to hide specific pixels on a matrix of a digital camera, such as mapping in the production of the matrix directly from the factory and dynamic mapping when using the camera. At the same time, the pixels marked as defective do not participate in the construction of the image. The program creates a map of the location of dead pixels, calculates their coordinates, and then subtracts so that they will not be noticeable in the final frames.

The lens is closed with an opaque shutter, which guarantees full opacity. Several shots are taken with the maximum possible exposure time. The appearance of points other than black in the photographs indicates the presence of “hot” pixels on the matrix in these places. Based on the described methodology by Managed infrastructure services, two algorithms are proposed that implement the authentication method of digital cameras.

In the first case, in order to detect specific pixels on images in which color components have been combined or pre-processed (in this case, there is no access to the original intensity and illumination values of the pixels of the camera matrix), it searches for all pixels whose values are the arithmetic average of the values of neighboring pixels in sense of 4-connectivity interpolation method.

Due to the fact that the criterion for searching for specific points is sensitive to the image content (false alarms of the criterion are possible for non-special pixels). For more details, visit the Managed LAN services /Outsourced IT Support Nottingham page.

Using the method on images presented in JPEG format reduces the accuracy of work due to the fact that the lossy compression algorithm used is based on the elimination of high frequencies in the image, which leads to image blurring and, consequently, an increase in false positives when particular points are detected.


Comments are closed.