FlowAtom: Atom-Based Evidence Aggregation for Multi-Label Website Fingerprinting

Chongru Fan; Wentao Huang; Wei Wang; Zhenquan Ding; Jinqiao Shi; Wei Cai; Zhiyu Hao

cs.LGSep 24, 2026

FlowAtom: Atom-Based Evidence Aggregation for Multi-Label Website Fingerprinting

Authors: Chongru Fan, Wentao Huang, Wei Wang, Zhenquan Ding, Jinqiao Shi, Wei Cai, Zhiyu Hao

Organizations: School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing, China · Zhongguancun Laboratory, Beijing, China

Abstract

Identifying the set of monitored websites in mixed encrypted traffic is challenging because an individual flow often provides only partial evidence of website identity. To address this challenge, we propose FlowAtom, which constructs shared prototypes, called Atoms, from flow representations without website labels. Specifically, FlowAtom pretrains a flow encoder on external unlabeled traffic and aggregates Atom responses across flows within each observation window into a fixed-dimensional, permutation-invariant representation for monitored website-set prediction. Across Direct HTTPS, Trojan, and VMess, FlowAtom achieves micro-F1 scores of 97.82%, 94.43%, and 93.92% in closed-world evaluation, respectively, and consistently outperforms the evaluated baselines in open-world evaluation on windows containing monitored visits. The code is available at https://github.com/aimafan123/FlowAtom.

Figures & tables

Figure 1: Different websites can generate similar flows; aggregating complementary evidence across flows in an observation window supports prediction of the monitored website set.

Figure 2: Overview of FlowAtom: flow representation pretraining, Atom construction, and window-level multi-label prediction.

	Direct HTTPS					Trojan					VMess
Method	$1$	$2$	$3$	$4$	$5$	$1$	$2$	$3$	$4$	$5$	$1$	$2$	$3$	$4$	$5$
ARES	$67.91$	$57.14$	$40.49$	$37.41$	$35.77$	$88.20$	$73.37$	$62.42$	$57.07$	$53.73$	$89.89$	$74.74$	$63.73$	$61.79$	$53.82$
BAPM	$57.57$	$34.21$	$25.30$	$21.05$	$15.66$	$64.17$	$34.33$	$33.61$	$32.39$	$23.73$	$65.26$	$37.11$	$31.71$	$30.95$	$24.12$
TMWF	$68.85$	$35.72$	$29.39$	$25.45$	$20.22$	$79.42$	$42.38$	$36.84$	$36.55$	$30.00$	$82.83$	$46.91$	$39.67$	$39.98$	$32.53$
Flow-DF	$75.92$	$80.20$	$78.82$	$78.66$	$81.72$	$82.29$	$79.62$	$77.05$	$81.30$	$82.07$	$82.56$	$82.49$	$82.21$	$84.23$	$80.10$
CAWF	$84.72$	$86.59$	$86.06$	$81.00$	$84.19$	$82.41$	$84.75$	$81.87$	$80.80$	$80.52$	$80.85$	$79.85$	$77.39$	$81.79$	$73.62$

Table 1: Closed-world micro-F1 (%) in three traffic scenarios, with

m\in\{1,2,3,4,5\}

distinct monitored websites per observation window. Bold and underlined values indicate the best and second-best results in each column, respectively.

Explore similar work

More Than Meets the Eye: A Semantics-Aware Traffic Augmentation Framework for Generalizable Website Fingerprinting
May 12, 2026Youquan Xian, Xueying Zeng, Lingjia Meng +6Accurate Traffic Forecasting Network Traffic
School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing, China · School of Computer Science and Engineering, Beihang University, Beijing, China · Zhongguancun Laboratory, Beijing, China +2
GETA: Generalized Encrypted Traffic Analysis
May 29, 2026Ransika Gunasekara, Rahat Masood, Salil KanhereNetwork Traffic Threat Detection
University of New South Wales (UNSW)
ResAware: Cross-Environment Website Fingerprinting via Resource-Privileged Distillation
Jun 16, 2026Chongru Fan, Wei Wang, Wentao Huang +5Fingerprint Web
Beijing University of Posts and Telecommunications Beijing, China · Zhongguancun Laboratory Beijing, China

May 12, 2026, cs.LGJ/K move · Enter open · S save

More Than Meets the Eye: A Semantics-Aware Traffic Augmentation Framework for Generalizable Website Fingerprinting

Youquan Xian, Xueying Zeng, Lingjia Meng, Lei Cui+5

School of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing, China · School of Computer Science and Engineering, Beihang University, Beijing, China · Zhongguancun Laboratory, Beijing, China+2

Accurate Traffic Forecasting Network Traffic Deep Semantic Representations Fingerprint Robust Cross-Domain Generalization Augmentation

Open paper AskPDF arXiv

Deep learning-based website fingerprinting has emerged as an effective technique for inferring the websites users visit. Although existing methods achieve strong performance on closed-world datasets, they often fail to generalize to real-world environments, especially under geographic and temporal shifts. This limitation fundamentally stems from the coupled effects of two key challenges: application-layer resource composition variability and observable feature instability induced by cross-layer encapsulation. Intertwined, these factors induce systematic shifts between underlying application semantics and observable traffic features. To address the above challenges, we propose SATA , a semantics-aware traffic augmentation framework. Specifically, SATA first performs application-layer semantic augmentation based on protocol rules, expanding the resource composition patterns within each flow and frame sequence patterns under protocol constraints. Based on these augmented frame sequences, we further introduce a cross-layer feature alignment mechanism via knowledge distillation. It aligns frame sequence with packet-length sequence features, enabling cross-layer feature alignment between enhanced semantics and observable sequences. Extensive experiments show that SATA successfully generates traffic patterns that are absent from the training set but genuinely exist in the test set, and significantly improves the performance of mainstream models across diverse and complex scenarios. In particular, in open-world settings, SATA improves ACC by 90.81% and AUROC by 48.37%. The source code of the prototype system is available at https://anonymous.4open.science/r/SATA-B6C2/.

Abstract

Explore similar work

More Than Meets the Eye: A Semantics-Aware Traffic Augmentation Framework for Generalizable Website Fingerprinting

GETA: Generalized Encrypted Traffic Analysis

ResAware: Cross-Environment Website Fingerprinting via Resource-Privileged Distillation