In this new age of technology, the face is increasingly becoming a digital passport for authenticating identity. With biometrics and multi- factor authentication becoming the norm, it is essential to pinpoint how facial recognition software actually works and look at its limitations and major challenges to its efficacy.
Facial recognition & how it works
While facial technology is not a new phenomenon, it is very much evolving and is on an inevitable rise. Facial recognition technology that utilises biometric data is the least intrusive and fastest method and works with the most obvious individual identifier – the human face.
Facial recognition is already in use in airports, border security and many law enforcement areas, and software has been developed for computer networks and automated bank tellers that use facial recognition for user verification purposes.
Unlike placing a hand on a reader or positioning the iris in front of a scanner, facial recognition systems have distinct advantages because of their non-contact process. It is predicted that biometric facial recognition technology will soon overtake fingerprint biometrics as the most popular form of user authentication.
Facial recognition works by analysing the characteristics of an individual’s face through a digital video camera. It includes measuring the overall facial structure, including distances between eyes, nose, mouth and jaw edges.
These measurements are retained in a database and are used as a comparison when a user stands before the camera. Each individual has over 80 nodal points, which are measured creating a numerical code, called a faceprint, representing the face in the database.
Some of the nodal points measured by facial recognition technology include:
• Distance between the eyes
• Width of the nose
• Depth of the eye sockets
• Shape of the cheekbones
• Length of the jaw line.
Biometric systems operate in a four-stage process, where the system will locate the user’s face and perform matches against the claimed identity in the facial database. The process involves:
• Capture – a physical or behavioural sample is captured by the system during enrolment.
• Extraction – a unique data is extracted from the sample and a template is created.
• Comparison – the template is then compared with a new sample.
• Matching – the system then decides if the features extracted from the new sample are matching or not, usually coming to a decision in less than five seconds.
While face recognition is a well-studied system in which several techniques have been implemented to address wider issues, there is a significant lack of research and proper experimental analysis in the use of plastic surgery as a new challenge.
Variations in pose, expression, illumination, ageing and disguise are considered as major challenges in face recognition and several techniques, including complex face recognition algorithms, have been developed to address these challenges. Plastic surgery, on the other hand, is perceptually considered as an arduous research issue.
Effects of plastic surgery on facial recognition
Plastic surgery is performed worldwide and is becoming prevalent due to advances in technology and the speed at which these procedures can be completed. For the first time on record, Americans spent more than US$13.5 billion on combined (surgical and non-surgical) aesthetic procedures in a single year (2015), according to statistics released by the American Society for Aesthetic Plastic Surgery (ASAPS).
James C. Grotting, MD, President of ASAPS stated: “Our industry’s growth is considerable, but not at all surprising. It reflects a healthy and robust economy wherein many people can afford to, and want to invest in themselves.”
“More people now perceive aesthetic enhancements and procedures as essential.” Plastic surgery is generally used for improving the facial appearance, for example, correcting disfiguring defects, removing birth marks, moles or scars or simply for rejuvenating an ageing face. However, misuse by individuals with the intent to commit fraud or invade law enforcement can lead to rejection of genuine users or acceptance of imposters.
The most prominent way that biometrics can be bypassed is through the use of cosmetic and plastic surgery. As a result, this can allow certain types of susceptible systems to return false positives when surgery tailors certain physical traits of a person to match those of which an algorithm searches. The inverse is also true. Individuals who undergo cosmetic surgery may not be recognised as themselves after the surgery.
To understand the challenges presented by cosmetic and plastic surgery as it applies to facial recognition, it is important to understand there are two types of results from cosmetic surgery – local and global. Local surgery alters one feature such as the nose, while global surgery alters the entire facial structure.
The prevalence of plastic surgery has introduced a new challenge to designing future face recognition systems. This is due to the sensitive nature of the process, privacy issues involved, and the fact that after surgery the geometric relationship between the facial features changes significantly and there is no technique to detect and measure these changes.
In a study, Effect of Plastic Surgery on Face Recognition by Richa Singh, Mayank Vatsa and Afzel Noore (2009), six recognition algorithms based on appearance, feature and texture (namely Principal Component Analysis, Fisher Discriminant Analysis, Geometric Features, Local Feature Analysis, Local Binary Pattern, and Neural Network Architecture) were chosen for evaluating facial recognition systems.
It was found that because face recognition algorithms generally rely on this information, any variation could affect the recognition performance. For example, the face recognition algorithms could not handle global facial plastic surgery such as skin resurfacing and full face lift because the facial features and texture are drastically altered after surgery.
Also, with major skin resurfacing or procedures to look younger, none of the algorithms were able to correctly classify the faces. Similarly, any change in the local facial regions such as nose, chin, eyelids, cheek, lips and forehead degrades the verification performance.
It was found that among different types of plastic surgery, otoplasty had the lowest effect on the performance of face recognition. This is because most of the algorithms do not include the ear region for recognition.
While there are many benefits to facial recognition technology, there are clear ethical, social and engineering challenges that need to be addressed. It is suggested that more research will be beneficial in order to design an optimal face recognition algorithm that can also account for the additional challenges due to facial plastic surgeries. AMP
