Our Pixilated Images Are Not Protected Enough

September 18/2016

It has become clear that any information put on the Internet isn’t protected from being intercepted. Moreover, the pixilated images are at a special risk. The computerized algorithm of identifying a pixilated image personification gains up to seventy percent accuracy.

The attackers can use Torch templates for neural networks to identify people even in Facebook images. The up-to-date methods of protecting pixilated images are claimed to be weak by Max Planck Institute when concerning one’s identity using a photo.

Moreover, in order to identify YouTube video ‘faces’, the researchers have taken the available photos and blurred them with the YouTube special protective tool. When these changed images have been given for identifying the personality to the neural system the accuracy of the result has been over sixty percents and even more when giving up to five attempts.

So, by training the neural system in this way, the results can be enhanced many times over. The researchers call Facebook and YouTube sources owners for improving their defensive measures not only from people interference but also from machines.

More articles