 |
| Fawkes Algorithm |
Tracking photos and searching for their owners online It has become an easy thing that anyone can do by using a tool or a site to search for photos and access important information about their owners even their accounts on social media, although this may be useful in some cases; For example, if you are asking how do I know if my photo is shared or used by someone else? But that can become harmful because just as you can search and track your photos, others can upload your photos and track you on social media, or even use them to train automated tools that can then be used to track you in real life!
What is facial recognition?
A facial Recognition System is a technology by which unique features in a person's face can be known so that the system can then search for a photo or group of photos of the same person among a database of thousands and possibly hundreds of thousands of photos and extract only the photos that belong to this person.
Facial recognition systems rely mostly on artificial intelligence and deep learning techniques so that anyone with average experience in the field can train a facial recognition model using deep learning algorithms available openly on the Internet, and within a short period, they will have a tool that can recognize Faces, so that if a photo of a particular person is provided, the tool will provide all photos of the same person from the database that was provided.
What is the importance of facial recognition systems?
Facial recognition tools are among the tools that can be useful in many cases, for example, through which we can know the name of the owner of the photo if we want to search for a specific person and only have a photo of him; The most prominent example of this is the police who use these programs to search for images taken from the surveillance cameras of criminals.
- These programs can also be useful in case we want to know who is using our photos on the Internet. Anyone can use our photos to create fake accounts on any social networking site on the web, which is why some sites have provided tools for people to track who uses their photos like in Facebook Profile Picture Shield.
What is the danger of the facial recognition system?
- Like nothing in this world; Facial recognition systems have several drawbacks, the most prominent of which is violating the privacy of image owners and using their images to train automated models without the permission of the image owner. The most famous example is ClearView, which has collected more than 3 billion photos of real people from the Internet without their permission, just by using social media crawling techniques and uploading photos of users who uploaded them voluntarily.
After that, this company trained models using the images it collected and then built a search engine that allows some parties to search for a specific person’s photo and get all available information about that person, in a flagrant violation of users’ privacy. It is also dangerous for these systems that they may expose you to online stalking and harassment, or what is known as cyberstalking.
What is the Fawkes Privacy Project?
- The Fawkes Project is a project that aims to protect the privacy of users from facial recognition systems that may misuse their photos for purposes they were not allowed to do as we mentioned in the previous paragraph. This project was carried out by a group of researchers in the department of The SAND Lab from the University of Chicago, USA.
- This project aims to manipulate images to deceive facial recognition systems so that if two images are presented to one person, one of them has been manipulated through the Fawkes program and the other has not been manipulated, the facial recognition tool will not be able to match the two images, that is, the tool will recognize them that each A picture of a different person!
How to hide identity from photos?
Fawkes is a sophisticated algorithm that poisons models that try to extract unique characteristics from your images so that they can identify your images, by modifying a set of points or pixels in your image in a process called Image Cloaking. These adjustments are not perceptible to the naked human eye, nor even to machines that are trained to recognize your images.
These in-image edits are used as Trojans to allow Fawkes' algorithms to poison a deep learning model when it tries to rehearse your images by giving it a distorted or distorted version of the facial features you have or are in the image. In this case, if a model is trained on an image of you that has been manipulated by the Fawkes algorithm; Then a picture of you is taken in a public place without your knowledge and submitted to the model, it will not be able to match the two pictures 100% in most cases [according to the researchers].
This algorithm has proven very successful and effective against a group of the most famous facial recognition models, such as the Microsoft Azure Face API model, the Amazon Recognition model, and the ⁺⁺Face model.
