A Quantum-Resistant and AI-Resilient Real-Time Keystroke Protection Framework With Blockchain-Backed Decentralized Identity

Abstract

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 (ϵ = 2.0, δ = 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× 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 HMACSHA256 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 < 10−9) and from 72.4% to 2.0% (CMU, p < 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 µs per-keystroke latency with <1.5% CPU overhead. Acoustic and electromagnetic side-channel defenses remain design-level considerations identified for future empirical validation.