Using artificial intelligence in anomaly detection to improve the efficiency of telecommunications networks

DOI: 10.31673/2412-9070.2025.025596

Abstract

Telecommunications networks are becoming increasingly dynamic and complex due to the huge amounts of data they process. As a result, detecting anomalous events in these networks is essential to ensure security and business continuity. Traditional rule-based anomaly detection methods are no longer effective in the rapidly changing telecommunications environment. Thus, artificial intelligence plays an important role in overcoming these shortcomings.
The article critically analyzes the role of artificial intelligence, in particular deep learning methods, in modern systems for detecting anomalies in telecommunication networks. The evolution of methods is considered - from early strategies to modern methods based on artificial intelligence.
In addition to improving the accuracy of anomaly detection, the use of artificial intelligence allows you to adapt to new types of threats in real time. Systems built on machine learning are able to update their models independently as new behavioral patterns emerge in the network, which significantly reduces the time required to respond to incidents. Particularly promising are architectures based on recurrent neural networks and auto-encoders, which work effectively with data sequences and can detect even weakly expressed or hidden anomalies. In addition, the combination of deep learning methods with big data processing mechanisms allows for the scalability of solutions required to process information in networks with millions of users. The integration of such intelligent systems into the telecommunications infrastructure not only increases the level of security, but also contributes to the construction of self-learning networks capable of operating effectively in a highly dynamic environment. Moreover, these technologies open up opportunities for predictive maintenance, which reduces costs and minimizes downtime for critical services.
The importance of using hybrid models to increase reliability and resilience is emphasized. This will allow telecom operators to proactively manage anomalies and optimize performance in a data-driven environment.

Keywords: Artificial intelligence, telecommunication networks, machine learning, deep learning, anomalies, convolutional neural network, scalability.