![](bg8.png)
deep learning approaches, and demonstrated that al-
though deep learning can produce superior perfor-
mance when provided with a large volume of data,
statistical algorithms such as ITAD can perform much
better and thus be preferred for small datasets. We ac-
curately replicated TypeNet on the Aalto desktop and
mobile datasets, achieving 0.93% and 6.77% EERs
respectively. Applying statistical algorithms on the
same datasets, we achieved EERs of 12.25% and
20.74% respectively. However, on a much smaller
dataset (the CU mobile), statistical algorithms sig-
nificantly outperform deep learning (4.3% and 1.3%
EERs for Scaled Manhattan and ITAD vs 19.18%
EER for TypeNet). Hence, when working with small
datasets, statistical algorithms remain the state-of-the-
art. Furthermore, the deep learning approach is still a
grey box and additional work is required to further ex-
plain its inner working (i.e., explainable AI), whereas
the statistical algorithms are clear and easy to under-
stand as they are based on statistical concepts, such as
mean, variance, and percentile.
Future work will be directed at quantifying the
size of a small dataset that should not be considered
for deep learning approach. We will also apply both
approaches on other public datasets, e.g., (Murphy
et al., 2017), (Sun et al., 2016). To further improve
performance, we will train the deep learning approach
with “semi-hard triplets” and “hard triplets” (Schroff
et al., 2015), and implement a weighted score fusion
for the ITAD metric. In addition to dataset sizes,
the nature/settings under which the datasets were col-
lected (e.g., typing behavior and/or content typed) can
also have an impact on performance. We plan to fur-
ther study these in future.
ACKNOWLEDGEMENTS
This work was supported by US NSF award TI-
2122746.
REFERENCES
Acien, A., Morales, A., Monaco, J. V., Vera-Rodriguez,
R., and Fierrez, J. (2021). TypeNet: Deep learning
keystroke biometrics. IEEE Transactions on Biomet-
rics, Behavior, and Identity Science.
Ayotte, B., Banavar, M., Hou, D., and Schuckers, S.
(2020). Fast free-text authentication via instance-
based keystroke dynamics. IEEE Transactions on
Biometrics, Behavior, and Identity Science, 2(4):377–
387.
Black, P. E. (2019). Manhattan distance. Available online
at: https://www.nist.gov/dads/HTML/ manhattanDis-
tance.html. Last Accessed: 2019-06-15.
Bromley, J., Guyon, I., LeCun, Y., S
¨
ackinger, E., and Shah,
R. (1993). Signature verification using a” siamese”
time delay neural network. Advances in neural infor-
mation processing systems, 6.
Deng, Y. and Zhong, Y. (2013). Keystroke dynamics user
authentication based on gaussian mixture model and
deep belief nets. International Scholarly Research No-
tices, 2013.
Dhakal, V., Feit, A. M., Kristensson, P. O., and Oulasvirta,
A. (2018). Observations on typing from 136 million
keystrokes. In Proceedings of the 2018 CHI Confer-
ence on Human Factors in Computing Systems, pages
1–12.
Gunetti, D. and Picardi, C. (2005). Keystroke analysis of
free text. ACM Transactions on Information and Sys-
tem Security (TISSEC), 8(3):312–347.
Killourhy, K. S. and Maxion, R. A. (2009). Comparing
anomaly-detection algorithms for keystroke dynam-
ics. In 2009 IEEE/IFIP International Conference
on Dependable Systems & Networks, pages 125–134.
IEEE.
Maheshwary, S., Ganguly, S., and Pudi, V. (2017). Deep
secure: A fast and simple neural network based ap-
proach for user authentication and identification via
keystroke dynamics. In IWAISe: First International
Workshop on Artificial Intelligence in Security, vol-
ume 59.
Murphy, C., Huang, J., Hou, D., and Schuckers, S. (2017).
Shared dataset on natural human-computer interaction
to support continuous authentication research. In 2017
IEEE International Joint Conference on Biometrics
(IJCB), pages 525–530. IEEE.
Palin, K., Feit, A. M., Kim, S., Kristensson, P. O., and
Oulasvirta, A. (2019). How do people type on mobile
devices? observations from a study with 37,000 vol-
unteers. In Proceedings of the 21st International Con-
ference on Human-Computer Interaction with Mobile
Devices and Services, pages 1–12.
Schroff, F., Kalenichenko, D., and Philbin, J. (2015).
Facenet: A unified embedding for face recognition
and clustering. In Proceedings of the IEEE conference
on computer vision and pattern recognition, pages
815–823.
Sun, Y., Ceker, H., and Upadhyaya, S. (2016). Shared
keystroke dataset for continuous authentication. In
2016 IEEE International Workshop on Information
Forensics and Security (WIFS), pages 1–6. IEEE.
Teh, P. S., Teoh, A. B. J., and Yue, S. (2013). A survey of
keystroke dynamics biometrics. The Scientific World
Journal, 2013.
Wahab, A. A., Hou, D., Schuckers, S., and Barbir, A.
(2021). Utilizing keystroke dynamics as additional
security measure to protect account recovery mech-
anism. In ICISSP, pages 33–42.
Zhong, Y., Deng, Y., and Jain, A. K. (2012). Keystroke
dynamics for user authentication. In 2012 IEEE com-
puter society conference on computer vision and pat-
tern recognition workshops, pages 117–123. IEEE.
ICPRAM 2023 - 12th International Conference on Pattern Recognition Applications and Methods
370