Template Ageing and Longitudinal Verification in Fixed-Text Keystroke Dynamics: A Subject-Disjoint Study Across Eight Weeks
Eight-week keystroke study finds fixed-text biometric templates age about 1.7 error points per week across four matchers.
A subject-disjoint study measured template ageing in fixed-text keystroke dynamics using 40 passwords typed four times in each weekly session for eight weeks. Researchers compared a scaled-Manhattan matcher, a gradient-boosted classifier, a TypeNet-style recurrent model, and a TypeFormer-style Transformer across enrolment-to-query gaps of zero to seven weeks. Equal error rates rose for every matcher, from 14.6–27.2% at no gap to 25.5–37.1% at seven weeks, about 1.7 percentage points per week. Baseline accuracy spanned 12.6 points, more than the 2.3-point spread in accumulated ageing, and ageing never reordered the matchers.
- Forty fixed passwords were typed four times weekly for eight weeks.
- Equal error rate rose from 14.6–27.2% at no gap to 25.5–37.1% at seven weeks.
- Error grew about 1.7 points per week for every matcher.
- Matcher accuracy differed more than ageing rate and rankings never changed.
- TypeNet-style M3 was least accurate but aged slower than scaled Manhattan.
Full article276 words · extracted from arxiv.org · click to collapse
Behavioural biometric templates are widely believed to degrade as the gap between enrolment and verification grows, but few studies measure this template ageing effect directly under controlled conditions. We collected a longitudinal dataset of 40 fixed passwords, each typed four times per weekly session over eight consecutive weeks. We compare a scaled-Manhattan matcher (M1), a gradient-boosted classifier (M2), a TypeNet-style recurrent embedding model (M3), and a TypeFormer-style Transformer (M4) under a 5-fold subject-disjoint protocol and a design that jointly varies mechanism and the enrolment-to-query gap, from 0 to 7 weeks. Template ageing proves large and systematic. Error increases monotonically with the gap for every mechanism, from an EER of 14.6-27.2% at a gap of zero to 25.5-37.1% at seven weeks, or 1.7% of decision error per week elapsed (p < 0.001). However, the choice of mechanism matters more than its rate of ageing. Baseline accuracy spans 12.6 percentage points across the four mechanisms, the degradation each accumulates over seven weeks spans only 2.3 points, and ageing never reorders them. A matcher can therefore be chosen on same-session accuracy, with ageing managed by re-enrolment scheduling rather than by matcher selection. The two properties are nonetheless distinct, as M3 is the least accurate mechanism yet ages significantly more slowly than M1 under every specification tested. Training randomness also matters differently by architecture, with 58% of the recurrent model's fold-to-fold variance attributable to seed noise against 19% for the Transformer. Because the smaller ageing-rate differences are sensitive to modelling choices, while the accuracy differences and the ageing effect are not, we recommend that comparative ageing-rate claims be supported by seed-level score fusion, independent replication, and an alternative outcome-model specification.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.29851