Conditions fixed before data collection: Do human-versus-machine keystroke detectors respond when their inputs are replayed and recorded through a software input path? (prototype v7)
Year: 2026
Type: article
Abstract: The measurement conditions of prototype v7, written and amended (A1-A7, B1-B17) before any value was taken, and deposited before the first capture. They fix the series, pairs, detectors, reading rules (two-sided equivalence with a margin of 0.10, Tango and Wilson intervals, missing-data bounds), predictions, random keys, the run order in day blocks and the limits of a study that replays human keystroke records (KUPA-KEYS) and machine schedules through one software injection path and two loggers on one Windows machine. From the production plan to the analysis this file is not changed (amendment B17); deviations during the captures are recorded in a separate file of the study repository. The file holds no keystroke data, no text and no output of a language model. It is works/conditions/prototype_v7.json at commit 7a8a9cddf4b1e7403f8f658cb97e077d9dae9e3a of the study repository; SHA-256 e5c21a4d5bcf8c5a3b6a2132c310f71b643d4907c6dcce706cefa3c723dbb91e. Amendments made after this deposit, if any, will be published as new versions with their reasons.
Authors: Miura, Mikoto, Harada, Yuusuke
Institution: Hiroshima University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225390211