A Quantum-Resistant and AI-Resilient Real-Time Keystroke Protection Framework With Blockchain-Backed Decentralized Identity
IEEE Access · 2026 · 14 · 57879-57894
DOI: 10.1109/access.2026.3680275 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Yeojin Jang und 2 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Regional Innovation System and Education (RISE) Program through Daejeon RISE Center, Ministry of Education (MOE) and Daejeon Metropolitan City, Republic of Korea,.
Abstract (Originalsprache)
Modern browser-based keylogging and AI-powered timing side-channel attacks demand new input protection systems, and the looming threat of quantum computing further compounds the challenge. We present a real-time keystroke encryption framework addressing three threat horizons: 1) browser-level attacks via ChaCha20-Poly1305 AEAD encryption, 2) timing-based inference attacks via a DP-inspired hybrid obfuscation defense combining calibrated Laplace noise ( $\epsilon =2.0$ , $\delta =10^{-5}$ ), temporal jittering, and dummy keystroke injection, and 3) quantum computing threats via hybrid classical+post-quantum cryptography (ML-KEM-768) with TVPD-HORS authentication achieving 2.7– $5\times $ faster verification than standard hash-based signatures. The system integrates Hyperledger Fabric for tamper-evident mapping management, W3C Decentralized Identifier (DID) support for self-sovereign identity, and optional zero-knowledge authentication to eliminate password transmission. Session keys are derived via HMAC-SHA256 incorporating blockchain transaction hashes, binding each session to immutable ledger state. Empirical evaluation under 5-fold stratified cross-validation on both a 20-user synthetic dataset and the 51-user CMU Keystroke Dynamics Dataset demonstrates that the DP-inspired defense reduces timing-based user-identification accuracy from 85.8% to 5.5% (synthetic, Random Forest, $p \lt 10^{-9}$ ) and from 72.4% to 2.0% (CMU, $p \lt 10^{-11}$ ), approaching the respective random-guess baselines of 5.0% and 2.0%. Macro-averaged F1-scores drop from 0.72–0.87 to 0.01–0.06 across all tested classifiers. The encryption pipeline achieves $22~\mu $ s per-keystroke latency with < 1.5% CPU overhead. Acoustic and electromagnetic side-channel defenses remain design-level considerations identified for future empirical validation.
Förderung
- Regional Innovation System and Education (RISE) Program through Daejeon RISE Center
- Ministry of Education (MOE) and Daejeon Metropolitan City, Republic of Korea,
Zitieren
APA
Jang, Y., Jang, H., & Lee, S. (2026). A Quantum-Resistant and AI-Resilient Real-Time Keystroke Protection Framework With Blockchain-Backed Decentralized Identity. IEEE Access, 14, 57879-57894. https://doi.org/10.1109/access.2026.3680275
Vancouver
Jang Y, Jang H, Lee S. A Quantum-Resistant and AI-Resilient Real-Time Keystroke Protection Framework With Blockchain-Backed Decentralized Identity. IEEE Access. 2026;14:57879-57894. doi:10.1109/access.2026.3680275
BibTeX
@article{Jang2026,
author = {Jang, Yeojin and Jang, Harin and Lee, Sangkeum},
title = {A Quantum-Resistant and AI-Resilient Real-Time Keystroke Protection Framework With Blockchain-Backed Decentralized Identity},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {57879--57894},
doi = {10.1109/access.2026.3680275}
}
EUVIMED