Face Recognition: State of the Art, Algorithmic Analysis and Legal Implications

Facial recognition technologies are becoming increasingly important in many areas mainly in security and surveillance, with the use of smartphone authentication and digital identity verification. This growing popularity is mainly due to its practicality, but it also raises a number of important ethical, legal and social issues. The aim of this thesis is to provide a comprehensive overview of the current state of facial recognition systems, combining a technical analysis of the main algorithms with a discussion of the legal implications. The thesis starts by introducing the basic principles behind face recognition—what it is, how it works, and why it's become such a common biometric method. Its contactless nature, for example, makes it ideal for public environments or digital systems that require quick and hygienic identification. From there, the paper highlights some real-life applications such as criminal identification, smart city infrastructure, and everyday consumer tech. The central part of the thesis focuses on a comparative analysis of two widely used systems: MTCNN (Multi-task Cascaded Convolutional Networks) and Microsoft’s Azure Face API. These two tools are examined in terms of how they work (architecture), how well they work (accuracy and performance), and how they might be used in practical situations. While MTCNN is based on a deep learning pipeline for detecting faces, Azure Face is a cloud-based platform that provides a complete face recognition service. I look at how they perform in different conditions, their response time, and even their cost-effectiveness. Another important topic covered is the vulnerability of face recognition systems. With reference to the document ‘Rethinking the Vulnerabilities of Face Recognition Systems: From a Practical Perspective’, we will analyse how facial recognition systems can be tricked or circumvented, thus talking about the problems these algorithms can have. We will look specifically at some examples given in the paper such as malicious attacks, spoofing methods and data set distortion. These problems pose a risk both to the accuracy of the technology but also to the privacy of the users who use it. Finally, we will deal with the legal and ethical aspect of facial recognition by reviewing the regulations in force in Italy and the European Union, especially referring to the GDPR. We will discuss how facial recognition challenges fundamental rights such as privacy, informed consent and freedom of expression. I also wanted to make a brief mention of the use of this technology in non-EU countries by discussing the situation in China. This thesis aims to analyse and argue the functioning of facial recognition technology, mainly by analysing the algorithms, on the basis of what has been analysed we then want to argue some of the dangers that these algorithms can bring. Finally, we want to discuss what protection tools we have at our disposal in Italy and in Europe, citing and arguing the regulations in force that protect us from possible risks.