Abstract :

The Rapid Adoption Of Encrypted Communication Protocols Has Improved Network Privacy And Security, But It Has Also Reduced The Visibility Available To Conventional Traffic Classification Systems. Network Operators Increasingly Need To Identify Applications And Traffic Behaviors Without Inspecting Payload Contents. This Paper Presents A Robust Open-set Classification Framework For Encrypted Network Traffic Under Packet Loss And Partial Observation. The Proposed Approach Treats Traffic Classification As An Open-set Problem In Which Previously Unseen Applications Must Be Distinguished From Known Traffic Classes Rather Than Being Forcibly Assigned To An Existing Category. The Framework Combines Flow-level Statistical Features, Packet-size And Timing Characteristics, Sequence-aware Representation Learning, And An Unknown-class Rejection Mechanism. To Model Realistic Network Conditions, Controlled Packet Loss, Truncated Flows, And Limited Observation Windows Are Introduced During Evaluation. A Confidence-based Rejection Layer Is Used To Separate Reliable Known-class Predictions From Uncertain Or Novel Traffic. The Overall Workflow Includes Traffic Collection, Preprocessing, Partial-observation Simulation, Feature Extraction, Robust Representation Learning, Open-set Decision Making, And Performance Analysis. The Framework Is Designed To Improve Reliability When Encrypted Flows Are Incomplete Or When Traffic Belonging To Classes Not Present During Training Appears During Deployment. The Study Also Discusses Privacy, Deployment, Class Imbalance, And The Limitations Of Metadata-based Traffic Analysis.

Published:

06-10-2026

Issue:

Vol. 26 No. 10 (2026)

Page Nos:

1 - 6

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite