Autonomous systems, such as HVAC systems in smart buildings, are vulnerable to cybersecurity threats, which is the focus of the paper titled ‘Thwarting Cybersecurity Attacks with Explainable Concept Drift.’

• The study introduces a Feature Drift Explanation (FDE) module aimed at identifying features that cause drift in machine learning models due to cyber attacks.
• The method utilizes Auto-encoders to find latent representations of data and measures divergence using Minkowski distance.
• Experimentation shows that FDE can successfully pinpoint up to 85.77% of the drifting features, aiding in the adaptation of Deep Learning models under drift conditions.
• The proposed technique offers a promising direction for enhancing security measures against attacks that manipulate sensor data.

This research contributes significantly to the arsenal of tools available for protecting autonomous systems from cyber threats, offering a path towards greater systemic resilience.

By enabling a better understanding of drifting features, this method opens up new possibilities for real-time threat detection and response within smart environments.