Cyberattacks are becoming increasingly sophisticated day by day. As a result, the Authentication of users is very important in a secure system. The authentication is performed by something that a real user knows, has, or is. The last one is called biometrics, which is most linked with fingerprint and face modalities. You can also authenticate a user by his or her behavior, called behavioral biometrics. Behavioral authentication based on user behavior patterns like keystroke dynamics and mouse interactions has become a promising technique to enhance web security. This paper argues that the long-term reliability dimension operationalized as the ability to maintain consistent performance in real-world environments with diverse conditions has not been thoroughly explored. In this paper, through a systematized review of multiple studies related to behavioral authentication, we examine the impact of behavioral drift, variability in environmental conditions and attacks on the precision and reliability of behavioral authentication systems. Additionally, from a network and internet perspective, issues such as packet loss, latency, and insecure communication channels can further challenge and affect real-time performance and data quality, reducing the reliability confidence. Also, other user's psychological factors can also affect and impact the user behavior. The outcome shows that single-modal techniques suffer a reduction of up to 15% in open environments, while multi-modal approaches (keystroke, mouse movements or other behavioral signals) maintain error rates below 2% consistently.
Behavioral authentication, keystroke dynamics, mouse dynamics, multi-modal fusion, reliability
© IACyC is conducted under the administration of the CyberMACS Erasmus-Mundus Masters Programme. cybermacs.org