SOLUTION OF SOCIAL PROBLEMS IN
MANAGEMENT AND ECONOMY
International scientific-online conference
88
A FEDERATED LEARNING APPROACH TO USER DATA PRIVACY IN
AI-DRIVEN TELECOM SERVICES
Khaydaraliyeva Khilola Farhod qizi
Tashkent University of Information Technologies
named after Muhammad al Khwarazmiy Assistent
Suyunov Shohjahon Xolmumin ugli
Tashkent University of Information Technologies
named after Muhammad al Khwarazmiy
3rd year student of the Faculty of Telecommunication Technologies
https://doi.org/10.5281/zenodo.15661211
Abstract
As artificial intelligence becomes central to telecom service optimization
and personalization, ensuring the privacy of user data is a growing challenge.
Traditional centralized machine learning methods require raw data aggregation,
exposing sensitive information and risking regulatory violations. This paper
presents a federated learning (FL) approach tailored for telecom environments,
enabling AI model training directly on distributed user devices without
transferring personal data to central servers. The proposed system integrates
differential privacy and secure aggregation mechanisms to enhance protection
while preserving model performance. Experimental evaluations using synthetic
mobile usage data demonstrate that our FL models achieve up to 96% of the
accuracy of centralized baselines, while significantly reducing privacy leakage
risks. The results confirm that federated learning is a scalable, privacy-
preserving solution for AI-driven telecom services that aligns with global data
protection standards.
Keywords:
Federated Learning, User Privacy, AI in Telecom, Secure Aggregation, GDPR
Compliance, Edge Intelligence, Differential Privacy
Introduction
The rise of artificial intelligence (AI) in the telecommunications sector has
enabled advanced features such as predictive network maintenance, intelligent
traffic routing, and personalized service delivery. These innovations rely heavily
on the analysis of vast amounts of user-generated data collected through mobile
applications, devices, and network logs. However, this data often contains
sensitive information such as location, usage behavior, and personal
preferences, raising significant concerns about user privacy and data protection.
SOLUTION OF SOCIAL PROBLEMS IN
MANAGEMENT AND ECONOMY
International scientific-online conference
89
Traditional AI development in telecom environments typically involves
centralized machine learning models, where raw user data is transmitted to a
central server for training. While effective in performance, this approach poses
major privacy risks and may violate data protection regulations such as the
General Data Protection Regulation (GDPR) in Europe, the California Consumer
Privacy Act (CCPA), and similar laws worldwide. Moreover, the growing public
demand for digital privacy and transparency has pressured telecom operators to
seek alternative, privacy-preserving solutions.
Federated Learning (FL) emerges as a promising paradigm to address these
challenges. Instead of centralizing raw data, FL enables distributed model
training across edge devices (e.g., smartphones, IoT nodes, or base stations),
where only encrypted model updates are shared with a central aggregator. This
decentralized approach not only enhances data privacy but also reduces the risk
of large-scale data breaches and lowers bandwidth consumption.
This study investigates the design and implementation of a federated
learning framework specifically tailored to AI-driven telecom services. We aim
to demonstrate that FL can maintain high model accuracy while significantly
improving user data privacy. To further strengthen the privacy guarantees, we
integrate differential privacy and secure aggregation techniques into the system
architecture. Through simulations on telecom usage data, we evaluate the trade-
offs between privacy, accuracy, and communication efficiency, highlighting the
practical viability of federated learning in future telecom infrastructures.
Results
The proposed federated learning (FL) framework was evaluated through
extensive simulations involving 1,000 distributed clients simulating mobile user
devices. The results demonstrate that FL can deliver high model performance
while preserving user privacy and ensuring communication efficiency.
3.1 Model Accuracy and Convergence
The FL model trained on distributed telecom usage data achieved a test
accuracy of
91.7%
, which is approximately
96%
of the performance of a
centralized model trained on the same data. The model converged in
28 global
training rounds
, demonstrating stable learning behavior under non-iid (non-
identically distributed) data conditions—common in telecom environments
where user behavior varies widely.
SOLUTION OF SOCIAL PROBLEMS IN
MANAGEMENT AND ECONOMY
International scientific-online conference
90
Model Type
Test
Accuracy
Rounds
to
Converge
Centralized
95.5%
—
Federated (no DP)
93.8%
25
Federated
+
DP
(ε=3)
91.7%
28
The framework scaled effectively to
10,000 simulated clients
, maintaining
stability in training accuracy and convergence. Variance across clients was
handled via
adaptive client sampling
, ensuring consistent contributions from a
representative subset of devices.
Conclusion
This study presents a federated learning (FL) framework tailored to the
unique privacy, performance, and scalability requirements of AI-driven telecom
services. By decentralizing model training and keeping user data on local
devices, the proposed approach effectively addresses key privacy concerns
while preserving high model accuracy. The integration of differential privacy
and secure aggregation further strengthens the system’s protection against
inference attacks and data leakage.
Simulation results confirm that FL can achieve over 90% model accuracy
compared to centralized baselines, with significantly reduced privacy risks and
manageable communication costs. These findings highlight FL’s practical
potential for enabling intelligent telecom services—such as service
recommendations and network optimization—without violating data protection
regulations or user trust.
Going forward, federated learning offers a sustainable and user-respecting
pathway for telecom operators as they transition toward 5G/6G networks, edge
intelligence, and regulatory compliance. Continued research into efficient model
architectures, adaptive communication strategies, and real-world deployment
scenarios will be essential to realizing the full benefits of FL in next-generation
telecommunications
References:
1.
H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas,
"Communication-efficient learning of deep networks from decentralized data,"
in Proc. 20th Int. Conf. Artificial Intelligence and Statistics (AISTATS), 2017, pp.
1273–1282.
SOLUTION OF SOCIAL PROBLEMS IN
MANAGEMENT AND ECONOMY
International scientific-online conference
91
2.
J. Konecny, H. B. McMahan, F. X. Yu, P. Richtarik, A. T. Suresh, and D. Bacon,
"Federated learning: Strategies for improving communication efficiency," arXiv
preprint arXiv:1610.05492, 2016.
3.
N. D. Lane, A. Bhattacharya, S. Georgiev, and P. K. Ravi, "An early resource
characterization of deep learning on wearables, smartphones and Internet-of-
Things devices," in Proc. 2015 Int. Conf. Internet of Things Design and
Implementation (IoTDI), 2015, pp. 1–10.
4.
R. Shokri and V. Shmatikov, "Privacy-preserving deep learning," in Proc.
22nd ACM SIGSAC Conf. Computer and Communications Security (CCS), 2015,
pp. 1310–1321.
5.
C. Dwork, A. Roth, "The algorithmic foundations of differential privacy,"
Foundations and Trends® in Theoretical Computer Science, vol. 9, no. 3–4, pp.
211–407, 2014.
6.
K. Bonawitz et al., "Practical secure aggregation for privacy-preserving
machine learning," in Proc. ACM SIGSAC Conf. Computer and Communications
Security (CCS), 2017, pp. 1175–1191.
7.
S. Wang, T. Tuor, T. Salonidis, K. Leung, C. Makaya, T. He, and K. Chan,
"Adaptive federated learning in resource constrained edge computing systems,"
IEEE Journal on Selected Areas in Communications, vol. 37, no. 6, pp. 1205–
1221, Jun. 2019.
8.
Google AI Blog, "Federated Learning: Collaborative Machine Learning
without Centralized Training Data," Apr. 2017. [Online]. Available:
https://ai.googleblog.com/2017/04/federated-learning-collaborative.html
9.
G. Zhu, D. Liu, Y. Du, C. You, J. Zhang, and K. Huang, "Towards federated
learning in 6G: A survey," IEEE Internet of Things Journal, vol. 8, no. 22, pp.
15784–15816, 2021.
10.
European Commission, "General Data Protection Regulation (GDPR),"
2016. [Online]. Available: https://gdpr.eu/
