Machine Learning-Driven Adaptive Authentication: Strengthening Cybersecurity against High-Volume Data Breaches

Nisher Ahmed (Westcliff University)
Md Emran Hossain (Westcliff University)
Zakir Hossain (California State University, Northridge)
Md Farhad Kabir (University of Southern California)
Iffat Sania Hossain (California State University, Camarillo)

Formosa Journal of Multidisciplinary Research (FJMR)
Vol. 4 No. 2 (2025): February 2025

As cyberattacks become more frequent and sophisticated, traditional static authentication methods have failed to protect them against it. And so given that some high-volume data breach incidents have highlighted vulnerabilities inherent in traditional username/password authentication, we must abandon these notions and embrace adaptive machine-learning (ML)-driven authentication systems that dynamically alter system security based upon real-time risk assessment. To strengthen cybersecurity resilience, this study introduces an ML-driven adaptive authentication approach, in which behavioral biometric, contextual information analysis, and anomaly detection algorithms are leveraged. We use a deep risk assessment methodology that dynamically re-authenticates logins based on device characteristics, geo location histories, behavioural analytics and historical user behaviour.