An Emirati researcher’s explainable AI technique detected 14 types of cyberattack with 97.45% accuracy while sharply reducing the computing power and time required.
Emirati data scientist and lecturer Mohammed Al Shehhi has developed an artificial-intelligence technique designed to spot cyberattacks targeting home surveillance cameras and other internet-connected devices, achieving reported accuracy of 97.45% and a marked reduction in detection time compared with an earlier international benchmark. According to Emarat Al Youm, the work points to a practical way of strengthening security for smart devices without relying on heavy computing resources.
Al Shehhi told the newspaper that the research is aimed at a growing problem: the spread of connected appliances, sensors and domestic cameras that often lack the memory or processing power needed for conventional security tools. His approach uses explainable AI to identify the most relevant indicators in network traffic, rather than examining a large set of variables, allowing the model to become smaller and faster while retaining a high level of precision.
The results were presented at the 12th World Internet of Things Conference in Abu Dhabi, held at Khalifa University from October 26 to 29, where academics, industry specialists and public-sector experts gathered to discuss artificial intelligence, cybersecurity and smart cities. The setting underlines the strategic relevance of the research in a field where the security of connected home devices has become a prominent concern.
In the tests described by Emarat Al Youm, the main model was able to distinguish normal traffic from 14 kinds of cyberattack. It recorded a prediction time of about 0.87 milliseconds on a small device, compared with roughly 27 milliseconds in the earlier reference study on the same hardware. That makes it about 31 times faster on the timing measure reported. The model was also reduced to around 301 kilobytes from 5,213 kilobytes for the original version used as the starting point, while a smaller variant was cut to about 152 kilobytes.
The broader research context supports the importance of that direction. A systematic review in Information found that machine-learning methods are increasingly used to detect malware in smart homes, with early detection, intrusion detection and stronger authentication among the key defences. A separate review in Computers & Security highlighted the need for better datasets that reflect the complexity of real smart-home environments, while a study in Sensors examined the specific risks posed by home security cameras and the need for automated traffic analysis. Together, those studies show why compact, fast and explainable models are attracting attention as smart homes become more common.
According to the report, the technique remains in the research phase and is not yet being used commercially in homes or industrial sites. Even so, its performance on standard datasets and on a real small device suggests a route towards lighter-weight protection systems for consumer electronics, household networks and, eventually, industrial Internet of Things deployments.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
