Facial recognition has come on in leaps and bounds in recent years. Initially the property of Hollywood Spy films and science fiction shows, facial recognition has become more and more ingrained into our society as, unsurprisingly, technology has evolved. This article seeks not to compare different recognition systems but present the case of advancements in facial recognition and seek to pose the questions of where we go from here.
But first, what is Facial Recognition technology? To sum-up neatly, it’s a form of biometric authentication. An image of the subjects (in this case a human) face is taken and then a series of algorithms are run that define the facial features – this can be defined in more granular terms;
- ‘Find’ the face in the image – using an algorithm that typically determines differences in pixel value. There is a plethora of methods in this step – the commonly used approach on cameras is the ‘Histrogram of Gradients’ where the difference between adjacent pixel values is represented as a series of arrows or ‘gradients’.
- Digitally transform the face into a consistent centred projection (to account for faces that are not in the ideal position/perspective in the image). Again there are many ways of approaching this but a common form is ‘Face Landmark Estimation’ where a series of specific facial ‘landmarks’ are trained into the algorithm. The algorithm is capable of then recognising these key points on any image.
- Use a pre-determined set of metrics to effectively measure and quantify the face in the image. This process is also known as ‘Embedding’ i.e. transforming a physical entity – in this case the image of a human – into a string of computer-generated numbers able to be used, stored and manipulated by a computer.
- Compare the metrics obtained from the ‘Embedding’ process of the imaged face to a database of metrics of known faces using a ‘Classification’ algorithm
As with many machine processes, the steps outlined above are analogous to those followed by the human brain – the emulation of this function is the ambition of neural networks – a class of Artifical Intelligence (AI) algorithms. This topic is extremely hot in modern technology as we are currently under-going the AI (or 4th) industrial revolution where many industries are prone to disruption by developments in the AI world – facial recognition is no stranger to this.
Facial recognition software is now becoming commonplace on smartphones – the Apple iPhoneX included ‘FaceID’ as the standard authentication protocol in-place of ‘TouchID’. What started out as being a security application and a mainstay of the ‘Mission Impossible’ film saga, could potentially turn into one of the greatest threats to personal privacy the world has ever seen. It is widely acknowledged that the average user would not be the target of sophisticated hackers looking to exploit the weakness within the ‘FaceID’ programming, but that senior executives or government officials could have their facial profiles stolen and utilised to access sensitive information. The traditional ‘Mission Impossible’ latex face mask will be consigned to history as hackers can access top-level secure documentation from the comfort of their bedrooms.
Does this make for a worrying future? As with all problems we as a society face, innovation will provide the answer and digital security experts have been developing innovative products unleashing a new wave of digital authentication. Therefore the question is not so much, is this something we should be concerned about? But more, what does the future hold? At least in the short-term, all we have to do to unlock our phone is stare at it.
Peter Clark, Technical Consultant, Leyton UK