This research introduces an end-to-end deep learning approach for detecting personally identifiable information (PII) in mobile network traffic without manual feature engineering.

• Uses triplet loss-based fine-tuning to improve network embedding quality
• Eliminates the need for external feature selection before model training
• Enables more effective identification of personal data exfiltration from mobile devices
• Strengthens security measures against unauthorized data leakage

This advancement matters because it provides a more robust framework for protecting user privacy by automatically detecting when sensitive information is exposed in network communications, addressing a critical cybersecurity challenge in mobile ecosystems.

End-to-End triplet loss based fine-tuning for network embedding in effective PII detection