A survey on safeguarding critical infrastructures: Attacks, AI security, and future directions

Technologies such as artificial intelligence (AI), blockchain, and the Internet of Things (IoT) have converged in driving the next wave of digital revolution. Amalgamating the aforementioned advancements with critical infrastructure (CI) can significantly help society by offering a quality of life and boosting the nation's economy and productivity. However, the lack of cybersecurity in CI gave rise to advanced threats and vulnerabilities that hindered the aforementioned societal benefits. In this vein, the paper provides an in-depth analysis of cyber threats and risks associated with different critical infrastructures, such as the financial, agriculture, energy, and healthcare sectors. Further, we thoroughly investigate the staggering benefits of AI and, based on it, present an exhaustive solution taxonomy to showcase the competency of AI mechanisms in confronting cyberattacks on CI. The taxonomy specifically addresses issues like data privacy, algorithmic bias, and human -AI collaboration for CI. Further, we proposed an AI-based secure data exchange framework for smart grid CI, where we attempt to secure the sensor's data (i.e., power consumption, energy readings, and network data) from malicious adversaries. The proposed framework is evaluated using statistical measures, such as accuracy, training time, and receiver operating characteristic (ROC) curve, and anomaly detection. Further, the paper examines the research challenges that still adhere to the critical systems and require stringent AI-based mechanisms to tackle them.