QuicReflo: Exploiting Multi-Flow Redundancy for Restoration-Free Covert Communication Over QUIC
Year: 2026
Type: article
Abstract: The rapid proliferation of HTTP/3 has consolidated QUIC as the backbone of modern web traffic, providing a ubiquitous and statistically diverse carrier for covert communication. Existing covert channels built on QUIC traffic predominantly manipulate packet-level patterns and depend on restoration at the receiver to reconcile flow inconsistencies. Such restoration, however, introduces traffic inconsistencies and exposes the channel to straightforward correlation attacks. Meanwhile, contemporary QUIC browsing typically involves multiple concurrent flows, whose complex spatio-temporal dynamics remain an underexplored opportunity for covert embedding. In this study, we propose QuicReflo, a restoration-free covert communication framework that operates across multiple QUIC flows. QuicReflo strategically schedules packet insertions guided by the backbone-and-adaptation redundancy modeling process, which predefines statistically admissible embedding regions to support lossless payload extraction without compromising observed traffic or legitimate communication. The framework integrates a redundancy-aware hierarchical diffusion model that precomputes redundancy space vectors, together with an embedding strategy that balances capacity, latency, and statistical fidelity. This joint design delineates embedding spaces that preserve the distributional integrity of QUIC flows while avoiding costly online alignment. Extensive experiments on real-world HTTP/3 traffic, collected from over 100,000 sessions across diverse QUIC websites, demonstrate that QuicReflo outperforms three state-of-the-art baselines. It achieves an average covert throughput of 28 bytes per packet while substantially constraining the detection accuracy of fourteen advanced classifiers to as low as 60.2%. By eliminating restoration and exploiting multi-flow redundancy, QuicReflo establishes a new direction for resilient and high-capacity covert communication over modern web protocols.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7201835842