article · Journal of Computer and Communications
MIES-TR is an intelligent model designed to detect syllable boundaries in real time as a user types on a keyboard. The system treats syllables as intermediate biometric units that capture both linguistic structure and stable motor behaviour, with the aim of improving continuous identity authentication. Its neural architecture uses character-position encoding, multi-scale convolutions, a unidirectional causal LSTM, and a sliding local attention mechanism. Operating entirely in a streaming setup without needing future input, the model achieves a processing latency under 30 milliseconds per keystroke. In evaluations using an annotated user corpus, the framework attained an average F1-score of 89.9 per cent and a word accuracy of 84.2 per cent. These findings show robust performance across different users, supporting the use of real-time typing dynamics for continuous security.
Traditional computer security often relies on one-time logins that cannot confirm who is sitting at the keyboard afterwards. By analysing the rhythm and structure of syllables as a person naturally types, this approach provides non-intrusive, continuous verification. It operates quickly enough to check identity in real time without interrupting the user or requiring special biometric scanning hardware.
The technology is applied and tested in an experimental setting, targeting continuous behavioural authentication and real-time cybersecurity systems. Potential end users include developers of adaptive access control frameworks, enterprise security providers, and interactive platforms requiring invisible identity checks. Future integration could enable deployment in embedded devices, multilingual applications, and multimodal biometric security, though further translation from corpus testing into live operational software is required.
AI-generated from the published abstract. Always read the original work before citing.
We introduce in this paper MIES-TR, an intelligent model for real-time syllable boundary detection during keyboard typing. This innovative approach positions the syllable as an intermediate biometric unit, combining linguistic richness and motor stability to enhance continuous authentication systems. MIES-TR is built around an optimized neural architecture consisting of character-position encoding, multi-scale convolutions, a unidirectional causal LSTM, and a sliding local attention mechanism. Unlike traditional offline syllabation methods, our model operates in a streaming fashion, without access to future input, and achieves a latency of less than 30 ms per keystroke, enabling dynamic, efficient segmentation compatible with interactive environments. Experimental results on an annotated user corpus demonstrate strong performance, with an average F1-score of 89.9%, word accuracy of 84.2%, and proven inter-user robustness, confirming the relevance of syllabic dynamics as a behavioral identity vector. Beyond accuracy, MIES-TR naturally integrates into adaptive security architectures such as ABAC policies enhanced with dynamic attributes, offering concrete prospects in free typing, multilingual support, multimodal biometric fusion, and embedded device implementation. MIES-TR thus paves the way toward smoother, invisible, and more robust authentication at the intersection of language processing, behavioral biometrics, and real-time cybersecurity.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.4236/jcc.2025.1312010
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
