Detecting LLM-Assisted Vietnamese Writing via Keystrokes under Behavioral Manipulation

2026-10-05

AI Summary

This study addresses the insufficient robustness of keystroke dynamics-based detection for LLM-assisted Vietnamese writing under behavioral manipulation. To this end, a dedicated keystroke dataset and a behavior-based threat model are constructed. Methodologically, the detection performance of temporal features and sequence modeling approaches, including 1D-CNN and TypeNet, is systematically evaluated, while adversarial training is introduced to generate controlled data variants that enhance generalization. Experimental results demonstrate that sequence models significantly outperform conventional feature-based methods. Furthermore, adversarial training effectively improves recognition rates for text-rewritten and behaviorally manipulated samples. These findings confirm the critical role of diverse writing behaviors in achieving detection generalization, substantially enhancing the overall robustness of the system against evasion attempts in real-world deployment scenarios.

Abstract

We study the robustness of keystroke dynamics for detecting large language model (LLM)-assisted writing. We introduce a Vietnamese keystroke dataset capturing realistic writing modes, including bona fide composition, transcription, and paraphrasing. We also define a behaviorally grounded threat model in which users deliberately alter typing patterns. To implement the threat model, we create behaviorally manipulated variants of the data designed to evade keystroke-based detection. We evaluate four keystroke modeling approaches: temporal and rhythmic representations, and sequential representations modeled with a one-dimensional convolutional neural network (1D-CNN) and TypeNet, under user-independent and context-independent settings. The results show that sequential models outperform feature-based approaches in most cases and that keystroke signals encode discriminative information about the writing process. However, detection is not uniformly robust: transcription is reliably identified, while paraphrasing and adversarially manipulated samples are frequently misclassified as bona fide when not explicitly modeled. To address this, we incorporate adversarial training using behaviorally manipulated data, which substantially improves separability and robustness. These results suggest that keystroke-based detection depends critically on exposure to diverse writing behaviors, and that strong performance under limited conditions does not generalize to realistic or adversarial settings without targeted modeling.

Problem

• keystroke dynamics
• LLM-assisted writing detection
• behavioral manipulation
• Vietnamese
• adversarial robustness

Innovation

• Keystroke Dynamics
• LLM-Assisted Writing Detection
• Behavioral Manipulation
• Adversarial Training
• Sequential Modeling