5 fashionable ways to hide from face identification
Today, the face recognition algorithm is used everywhere. And not only in professional tracking systems or smart web services, but in various entertainment applications and basic functions like autofocus. Poster Daily decided to try to trick the algorithm with makeup.
Different systems for recognizing a human face have different goals, which is why their algorithms differ. One of the first systems that were able to establish a person’s identity from a photo looked for nodal points on the image that made up the main features of the face, and measured the distance between them. The main measurement parameters are the distance between the eyes, the width of the nose, the depth of the orbit, the shape of the cheekbones, and the length of the jaw line. Key figures in digital terms constituted a unique code or faceprint representing a person in the database. So to establish the identity of the suspect, his photo is uploaded to the system, and the faceprint is checked with thousands of other similar ones.
The problem with the method is that for different shooting angles, lighting, and even facial expressions, the measured values can vary. Therefore, more modern programs use 3D-modeling of faces, which gives more accurate results. By analyzing portions of the face in the photograph where the hard tissues and bones are more obvious – the eye sockets, nose, chin (these areas are unique and do not change over time), the program builds a three-dimensional model of the face and after that “unfolds” the model as you like – in profile or full face, in parallel measures facial features on a submillimeter scale.
But this method is not ideal – for a 100% result, it is desirable to check and identify several photos.
For the most accurate indicators today, biometric algorithms are used that analyze unique skin textures, irises, or even veins.
The process of analyzing a surface texture works essentially like face recognition. The image of a person (for example, his face) is divided into smaller blocks, and algorithms distinguish the smallest lines, wrinkles, pores, moles and even skin structure on the texture and iris of the eye – all this appears in the form of a mathematical model. Thanks to this system, it is easy to distinguish even between identical twins.
The simplest face analysis algorithms are those used in entertainment applications like Snapchat, as well as in photo programs where the focus is automatically set on the face. These programs analyze the oval of the face, as well as the position of the main features of the face – eyes, nose, mouth.
DETAILS BY TOPIC
8 tools for cyber paranoids instead of a foil cap
How make-up camouflage works
The very first stage of face recognition in any system is face detection in the image. When the program does not see the face, then it does not make sense either to analyze facial features or to calculate the distances between key points. So it’s logical to block the system at this stage. The face search in the frame works according to a rather simple algorithm that analyzes the presence of an oval of the face, eyes, nose and mouth. However, with the help of a simple set of cosmetics, face patterns can be broken, making the car think – it can be anything, but not a face.
5 main principles of make-up camouflage
Unfortunately, the drawings on the face will not help to hide from the identifier. Just a picture on the face, even if it is not made in the classic color scheme, is still doomed to the ears of a Snapchat dog. The first thing the identifier is guided by is the oval of the face, ears and nose. Only tape can help here, which will change outlines completely. Glue the nose to the cheek, and roll the ears into tubes, distribute makeup, use unusual tones and techniques so that there is not much difference between the color of the eyes, cheeks and lips. Do not emphasize the eyes or cheekbones – on the contrary, it will simplify face recognition for the system.
Two symmetrical eyes are a clear sign of a face for computer vision. Try to hide both eyes or at least one of them. It is best to use materials that reflect light — sparkles, mirror pieces, or completely smooth, light-reflecting surfaces. Glare will distract the system from natural facial shadows and will not allow to recognize the face.
Try playing with the elliptical shape of the head and the symmetry of the ears. Unfortunately, this does not always help, since the system recognizes an oval shape and, if it finds shadows from the eyelids or nose, it works. The area where the nose, eyes and forehead are located is key to the face recognition system. Try to break this triangle by applying an unexpected element to the bridge of the nose, for example, bright paint (in our case, yellow).
Change contrasting tonal gradients, change the spatial ratio of dark and light areas of the face, using makeup, accessories and your hair. Suddenly sticking out curled locks of not a natural color (or rather several colors) and a change in the texture of the skin, stage makeup, the feeling of a molten face and the lack of color focus on the main details of the face.