Implementing artificial intelligence models for malware detection in critical telecommunications infrastructures

Abstract

The rapid advance of digitization and global interconnection has made critical telecommunications infrastructure more vulnerable to various cyber threats, among which, malware is one of the most persistent and damaging. In this context, implementing Artificial Intelligence (AI) models to detect malicious software early on has become a key strategy for ensuring the resilience, availability, and continuity of essential services. This article analyzes the primary machine learning and deep learning approaches used for malware identification, focusing on their implementation in critical telecommunications environments.

Classical techniques such as Support Vector Machines (SVM) and Random Forest are reviewed, as well as advanced neural network architectures, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Graph Neural Networks (GNN). The methodology focuses on a systematic review of recent literature and a comparative analysis of models based on performance metrics, generalization capability, and computational resource consumption. Results indicate that hybrid models combing deep learning and network flow analysis techniques canvachieve 98% rates detection with a significantly fewer false positive. Finally, current challenges and future perspectives for the integration of AI into the cyber defense of critical infrastructures are discussed, highlighting the importance of explainability, unknown threat detection, and collaboration between Telcos carriers and research institutions.