Research in the International Journal of Intelligent Engineering Informatics proposes the use of system behavioural modelling and unattended or semi-supervised machine learning to help solve the problem of cyber security in smart cities. By training machine learning models on relevant datasets, the researchers suggest that security systems can be improved so that they can identify and mitigate cyber threats. An ongoing challenge is to ensure the reliability and completeness of those very datasets to allow for anomalies to be detected with confidence.
The development of technologically connected and enabled ‘smart cities’ could help us face better rapid urbanization and growing populations. Connectivity allows Internet of Things (IoT) systems to be used more effectively and to harness the power big data analytics to tackle various urban issues, such as traffic congestion, air pollution, water management, housing issues, urban planning, healthcare, and equitable accessibility to resources for everyone. However, as with any integrated and networked technology, IoT devices can be vulnerable to unauthorized access by those with malicious intent. This is most worrying in the area of safety systems, but also of concern across many others such as transport and healthcare.
N. Girubagari and T.N. Ravi of the Thanthai Periyar Government Arts and Science College in Tiruchirappalli, Tamil Nadu, India, point out that cyber attacks in this context go way beyond the kind of the privacy concerns of individuals. They might affect smart infrastructure, communications, and e-governance. As such, intelligent detection systems based on machine learning to safeguard against cyber threats are needed urgently.
The team looks at various anomaly detection methods and assesses their pros and cons. The paper highlights existing obstacles and gaps in research that are currently stymieing the full potential of the smart city. The work evaluates and contrasts methods for identifying anomalies in big data-based cybersecurity, utilizing survival analysis to assess the benefits and drawbacks of current techniques. The long-term objective is, of course, the efficient detection of cyber attacks in real-world scenarios. The research emphasizes that performance assessment for machine learning methodologies is important at this juncture.
In future work, the researchers hope to conduct additional experiments to test performance and establish a methodology for precise and comprehensive anomaly identification in smart city systems.
Girubagari, N. and Ravi, T.N. (2023) ‘Methods of anomaly detection for the prevention and detection of cyber attacks’, Int. J. Intelligent Engineering Informatics, Vol. 11, No. 4, pp.299–316.