Table 1.
Summary of salient typing demand, analysis mode, and accuracy rates for a spectrum of different keystroke biometric approaches.
| Typing input demand | Method of analysis | Accuracy | Reference |
|---|---|---|---|
| 10 character string input 10 times with 30 participants | Statistical ( and ρ) | FAR = 1.89% | Araújo et al. (2005) |
| FRR = 1.45% | |||
| Circa 40 character string input 10 times with 100 participants | Statistical (GPD fused with DSM) | EER ≈ 1% | Teh et al. (2011) |
| Circa 30 character string input 10 times with eight participants | Statistical (GMM) | FAR = 2.1%; FRR = 2.4%EER < 3% | Hosseinzadeh et al. (2006) |
| Short phrase entry with six participants | Artificial Neural Net | Accuracy = 97.8% | Obaidat and Macchairolo (1993) |
| 15 valid and 15 invalid users × 225 sequence | ANN + Fuzzy logic | EER = 0% | Obaidat and Sadoun (1997) |
| Short password entered three times with 90 valid and 61 imposter participants | Multilayer back propagated ANN | FAR = 1.1%; FRR = 0%* | Lin (1997) |
| 7 character string input between 150 and 400 times with 25 participants | ANN using multilayer perceptron | FAR = 0%; FRR = 1% | Cho et al. (2000) |
| At least 8 character string input 25 times with 29 participants to study | Fuzzy logic | FAR = 2.79%; FRR = 7.379% | De Ru and Eloff (1997) |
| 683 character string using 154 participants | Statistical - trigraph-based | FRR = 4%; FAR = 0.01% | Bergadano et al. (2002) |
| Short (n < 15 characters) strings input 10 times with 23 participants | Auto-regressive classifier linked to pressure data | EER ≈ 3% | Eltahir et al. (2008) |
| 10 character password input to database enrolment with 50 samples (30 genuine and 20 forged) | Statistical; & Statistical augmented with pressure data | EER = 2.04%; EER = 1.41% (P-augmented) | Lv and Wang (2006) |
| 8 character string with 10 timing- and 10 pressure vectors recorded | Artificial Neural Net augmented with pressure data | EER values of 16.5, 14.94, and 11.78% for respectively, pressure, latency, and pressure + latency | Loy et al. (2007) |
| Short string pairs input 15 times with 20 participants | Independent component analysis and fast-ANN augmented with acoustic record | FAR = 4.12%; FRR = 5.55% | Nguyen et al. (2010) |
*
With refined thresholding.