How Can AI Turn Our Digital Privacy into a Myth?
Have you ever thought that maybe artificial intelligence can turn our digital privacy into a myth? One person thinks so, let's find out why.
Join the DZone community and get the full member experience.Join For Free
Machine Learning is revolutionizing every niche that it’s encountering, be it finance, healthcare, website development or even digital security. You must have seen videos on Youtube or posts on your news feed in which certain texts or a person’s face is blurred. Well, that’s how our digital privacy is ensured by simplest of technologies.
But think about it, in an age of Machine Learning, can’t your digital privacy be easily breached? The answer is a big "Yes," and a team of researchers at the University of Texas has proven that. They have developed a software that can identify the sensitive content hidden behind blurred or pixelated images. The content can be someone’s house or vehicle number, or simply a human face.
Interestingly, the team hasn’t used some state of the art technology to do it. It has instead used Machine Learning methods to train the neural networks. So instead of being programmed, the computer has been fed with large volumes of sample images. The algorithm used doesn’t actually unblur or restore the image. It identifies the content of the blurred image based on the information it already has.
You can refer to the image below, to see how AI reconstructs a blurred image.
Image Credit: Google Brain
You Must Be Wondering How Exactly It Happens? Well, Here’s How:
Firstly, the team feeds neural networks with data sets of images. As a neural network sees more words, faces, and objects, it gets better at its recognition skills. Once it achieves a good accuracy in identifying objects, say 90 percent, it is fed with blurred versions of the images used. This is how neural networks learn to distinguish or correlate between the blurred and original images.
Once the learning process is complete, the neural networks are exposed to entirely new images (Pixelated). This is the last step to see how accurately the software can recognize or reconstruct the given image.
You can refer to the image below, to see how neural networks reconstruct different objects.
Image Credit: Google Brain
Now let’s compare this technology with the current crop of image-enhancement tools and techniques. Such tools do sharpen images to some extent, but due to their emphasis on reconstructing the original image with pixel-perfect accuracy, they end up producing ‘Overly-Smoothened’ images that appear unrealistic. On the other hand, a neural network analyzes different regions of the original image and uses the semantic information to create realistic textures.
How Should We Look at This Development?
This technology has both an upside and downside. It can be incorporated into image-editing software like Photoshop or into operating systems so that the quality of images are restored while zooming. It can also augment the security measures by providing high-resolution images of suspicious vehicles to concerned organizations.
On the other hand, Machine Learning will be changing the dynamics of digital security and privacy, and we must be prepared for that. As the influence of Machine Learning will grow, its exploiters will increase as well. Therefore, developments like these should ring a bell among people who want to develop future technologies in the field of privacy. Such people must realize that no technology is foolproof, and it should be enhanced by doing a continuous analysis.
It’s high time we modernized the privacy-preserving methods before the Machine Learning techniques turn the very idea of digital privacy into a myth.
Opinions expressed by DZone contributors are their own.