TRIDENT -- A Three-Tier Privacy-Preserving Propaganda Detection Model in Mobile Networks using Transformers, Adversarial Learning, and Differential Privacy

2025-06-05

AI Summary

Balancing privacy preservation and model accuracy remains challenging in mobile network promotional content detection. Method: This paper proposes a three-tier collaborative defense framework: (1) a fine-grained semantic modeling layer based on Transformer architectures (BERT/GPT-2); (2) an adversarial training layer to enhance robustness; and (3) a novel syntax-level differential privacy mechanism integrating multilingual back-translation, character-level noise injection, and entity anonymization—perturbing both inputs and label distributions at the syntactic level. The decoupled architecture enables lightweight deployment on mobile devices while supporting controllable privacy–utility trade-offs. Results: The baseline achieves F1 scores of 0.89–0.90; with full three-tier protection, F1 drops moderately to 0.83, yet privacy leakage is significantly reduced and practical detection accuracy is retained—demonstrating the feasibility of simultaneously achieving strong privacy guarantees and high detection performance.

Technology Category

• Machine Learning: Privacy
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Computer Vision: Bias, Fairness & Privacy

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Security and Privacy: Privacy-enhancing technologies
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The proliferation of propaganda on mobile platforms raises critical concerns around detection accuracy and user privacy. To address this, we propose TRIDENT - a three-tier propaganda detection model implementing transformers, adversarial learning, and differential privacy which integrates syntactic obfuscation and label perturbation to mitigate privacy leakage while maintaining propaganda detection accuracy. TRIDENT leverages multilingual back-translation to introduce semantic variance, character-level noise, and entity obfuscation for differential privacy enforcement, and combines these techniques into a unified defense mechanism. Using a binary propaganda classification dataset, baseline transformer models (BERT, GPT-2) we achieved F1 scores of 0.89 and 0.90. Applying TRIDENT's third-tier defense yields a reduced but effective cumulative F1 of 0.83, demonstrating strong privacy protection across mobile ML deployments with minimal degradation.

Problem

• Detect propaganda in mobile networks accurately and privately
• Prevent privacy leakage while maintaining detection performance
• Combine multiple techniques for robust privacy protection

Innovation

• Transformers for propaganda detection accuracy
• Adversarial learning with differential privacy
• Multilingual back-translation for semantic variance