What is PhotoDetective?
Our current project is an application that enables users to determine if, and how, an image may have been altered. We have created a simple-to-use Graphical User Interface (GUI) so that anybody can take any image they find in commonly used formats (jpeg, png) and perform digital forensic analysis with only a point and a click of the mouse.
What can PhotoDetective do?
PhotoDetective allows users to open an image file and then select a forensic algorithm to perform on the image. These algorithms come from peer reviewed papers in the digital forensics field. Each algorithm tells a different part of the story behind the photo including:
- Is the lighting direction consistent?
- Was the entire shot at the same focal length?
- Have some parts of the image been compressed more than others?
What color filter array did the camera use to shoot the picture?
- What quality level was the image saved at?
- What quantization table was the image saved with?
- Has an image been recolored?
Are the jpeg artifacts distributed equally across the image?
- Is an area naturally or artificially blurred?
- What colors are the same in image?
- Does the image contain a hidden thumbnail of itself?
- Has the contrast been enhanced?
- What is the Exif Information?
- Is the camera clipping image colors?
*Actual* Examples of Output Saved Directly from the Program:
Photograph credits: 1. Justin Hoch photographing for Hudson Union Society; 5a. Sardaka 5b. Asdrubol
How does PhotoDetective work?
- Principal Component Analysis to detect consistency among artifacts in the image
- Wavelet decomposition analysis to detect whether the image is all at the same focal length
- Edge detection routines to identify naturally blurred vs. artificially blurred areas (used to covered up tampering)
- Error level analysis to detect whether parts of the image come from sources with different error potential (i.e., two different images)
- Noise enhancement to make small artifacts more visible
- Local Min/Max pixel highlighting to determine if the pixels errors display a normal variation
- Lighting gradient coloration to visually show where the light is coming from in an image, and is it consistently in one direction.
- Temperature Pseudocolor based on pixel intensity to identify odd or inconsistent coloration
- Pixel color randomization to show pixels that have the same RGB value more easily
- Extreme color highlighting to show where the pixels are pure white and black and if the colors are being clipped by the camera
- Thumbnail extraction from image to show what the picture might have originally looked like before manipulation
- Displaying Exif information to display properties concerning the image's history Extraction of Quantization tables used to save the image
- Minimum-Medium-Maximum RGB value for each pixel to identify coloration inconsistencies Save quality estimation to know what amount of compression has been used on the image.
- Display the histogram of colors to see if a contrast filter has been applied
- Color Filter Array Estimation to know what color filter arrangement the specific camera uses to interpolate the colors
What can I do with PhotoDetective?
Though we imagine that the primary users of our product would be concerned with criminal and legal matters, PhotoDetective has many real world applications that would be of interest to far more people. Images of the human body in advertising are commonly edited to present a completely unrealistic standard of beauty; much of this editing can be exposed using our tool. Digital archivists can learn more about the authenticity of images. Consumers hesitant about purchasing a product can now be more informed if preview images have been altered to hide faults and defects. Insurance agencies interested in determining whether a claim is fraudulent can do so at the press of a button.
At the present, there is nothing on the consumer market that provides a similar level of functionality. The interface was designed to be as intuitive and user-friendly as possible, while the coding was designed to be up to professional standards. We are grateful for support and are hoping this can be the beginning of many new things in the field of image forensics. Hope you will think of supporting us and will find our program useful!
Risks and challenges
It is important to remember that this is a forensic tool meant to aid your own common sense. PhotoDetective does not provide a yes/no answer - instead you get a picture which can be interpreted, along with a guide to show what sort of image represents what sort of alteration or tampering. We provide a variety of tools since different types of editing can be detected best using different algorithms.
This software part of the project is in the advanced stages and almost ready to be delivered. All of the images shown on this page come directly from the actual program. While most Kickstarters may take years if ever to deliver the final product, we offer a turnaround of just a few months (March at the latest).
During this time, we will be completing beta-testing, writing the user-guide, and putting the finishing touches on the program. We also offer a reward to those who simply can't wait and those select few will receive the program and guide, and they will also receive any updates or patches that are made before the final release.Learn about accountability on Kickstarter
- (35 days)