Intelligent access control system based on face recognition and user behavior analysis using deep learning methods

Today, access control systems that rely only on passwords and PIN codes can no longer keep
up with growing security requirements. These traditional methods are easy to steal, guess, or share
with someone else, which turns them into the weakest link in the protection chain. In this context,
solutions that use biometrics and can also “understand” user behavior, rather than just verify entered
data, are becoming especially relevant.
This study develops a software system that makes decisions not only based on a person’s
biometric features, but also on how they behave when trying to log in. The aim is to obtain a tool that
can detect suspicious activity in time and automatically strengthen access restrictions when necessary.
As a technical foundation, computer vision and deep learning methods are used to recognize
faces in a video stream. In addition, a behavioral analysis module is implemented that tracks the
sequence and frequency of login attempts, as well as the time and context of access. The system is
structured into several interconnected modules: face recognition, time- and zone-based access
verification, and risk level assessment.
The prototype shows that combining biometric and behavioral features can significantly
improve the quality of access control. When the number of errors or atypical actions exceeds a
predefined threshold, the system automatically blocks the user and triggers an alarm scenario. In this
way, the probability of unauthorized access is reduced, while part of the routine work of the security
administrator is offloaded to the system itself.
The scientific novelty of the proposed approach lies in integrating face recognition and
behavioral analysis not as two separate layers of protection, but as components of a single
decision-making architecture. This makes it possible to consider both the user’s biometric
characteristics and the context of their actions-such as time, location, and patterns of failed attemptssimultaneously
when deciding whether to grant access. Compared to classical single-factor solutions,
such an integrated approach forms a more flexible and intelligent security perimeter