Let's talk science for a bit shall we?

source
Imagine you're a lawyer defending a case in court and a digital image was filled as evidenced against your client. How would you verify the authenticity of that image?

That, my friend, is where Image Forensics comes in.
This problem of image authentication has been studied quite alot over the years and a lot of methods have been developed. One of such methods is to determine which camera was responsible for that image, or better yet to prove that a particular camera is responsible for the image. And this is what is referred to as image source identification.

Fascinating right? I thought so too.

I recently did a little review of some methodologies for Image Source Identification and i gotta say i was pretty intrigued. So much so I've decided to base my Masters Thesis on the subject.

Methodologies

Firstly, let me briefly describe the image formation process in digital cameras.

Source: Bayram, S., Sencar, H. T., & Memon, N. (2008). Classification of digital camera-models based on demosaicing artifacts. Digital Investigation, 5(1–2), 49–59. https://doi.org/10.1016/j.diin.2008.06.004

When a photo is taken, light enters the lens and is passed through a series of filters, the most important been an anti-aliasing filter. This light is then focused onto an array of Charge-coupled devices (CCD) or pixels in lay man's terms. The CCD obtains the electric signal representation of the photo been taken. all this takes place in the sensor masked with another filter called the Color Filter Array.
Each pixel in an image contains three components, Red,, Green, Blue which require separate CCD but due to the cost only one CCD is applied per pixel which captures only one component. The values for the other two channels are then Interpolated (guessed) using a Demosaicing algorithm based off of the pixels around it.
After demosaicing, other processes such as Gamma Correction and Noise Reduction are caried out and the image is saved to memory.

Now, in my study i discovered two major methodologies.

Due to imperfections of some of the materials used in manufacturing sensors, certain noise patterns are introduced into an image. One of which is the Photo Response Non-Uniformity (PRNU) which is unique across even different models of the same brand. It is sometimes referred to as a finger print. This can be used to determine the source of an image.
During Interpolation, Demosaicing algorithms introduce correlations between the different color channels in order to correctly interpolate the color components. This leaves certain traces on the images which can be used as a way to determine the source of an image.

Now This is just a very brief summary of these methodologies and i am looking to do some more research in the area of using PRNU based methodologies. Most methods i have seen use Unsupervised learning algorithms like Correlation Clustering and Support Vector Machines.. So i want to try something simple like Logistic regression.

I sincerely hope this wasn't a boring read for you. At least you've learnt something new.
Thank you for Reading.

Congratulations! This post has been upvoted from the communal account, @minnowsupport, by jo5h from the Minnow Support Project. It's a witness project run by aggroed, ausbitbank, teamsteem, theprophet0, someguy123, neoxian, followbtcnews, and netuoso. The goal is to help Steemit grow by supporting Minnows. Please find us at the Peace, Abundance, and Liberty Network (PALnet) Discord Channel. It's a completely public and open space to all members of the Steemit community who voluntarily choose to be there.

If you would like to delegate to the Minnow Support Project you can do so by clicking on the following links: 50SP, 100SP, 250SP, 500SP, 1000SP, 5000SP.
Be sure to leave at least 50SP undelegated on your account.

Image source identification.

Methodologies