The paper fuses existing research with the contemporary debates concerning SCADA and cybersecurity augmented with machine learning. Cyber threats predominantly target SCADA systems that find application in critical infrastructure. Ordinary security tools often fail to spot anomalies in real time and thus are unable to respond quickly enough to mitigate system vulnerabilities. The infusion of machine learning into cybersecurity frameworks, however, has proved worthy during threat detection in terms of accuracy and response efficiency. Continued Internet traffic monitoring of SCADA for potential threats is needed for SCADA cybersecurity. Based on their ability to detect deviations from usual system behavior, machine learning-based anomaly detection methods can apply a proactive methodology. This paper aims, first and foremost, to examine existing ML approaches and algorithms in this field. This provides an exhaustive survey of the anomaly detection area propelled by machine learning suitable solutions thereby outlining their pros and cons in softening cyber risks.