Abu Dhabi, UAE – An Emirati data scientist has developed a lightweight artificial intelligence model that can detect cyberattacks on internet-connected devices around 31 times faster than a previous benchmark, while maintaining accuracy of about 97.45%.
Mohammed Al Shehhi, a data scientist and lecturer at the Higher Colleges of Technology, presented the research at the 12th World Internet of Things Congress, held at Khalifa University in Abu Dhabi from 26 to 29 October 2026.
How it works
The model protects Internet of Things (IoT) devices such as cameras, smart appliances, sensors and industrial equipment. It uses explainable AI to identify the most useful network indicators, cutting the features used by one model from 47 to 20 without losing performance.
Performance at a glance
- Detects 14 types of cyberattack with around 97.45% accuracy
- About 0.87 milliseconds per detection – roughly 31 times faster than the previous benchmark on the same device
- Main model size of about 301 KB, compared with 5,213 KB for the full starting model
- A smaller version of about 152 KB runs in around 0.61 milliseconds
- Scored 0.969 on one key metric, versus 0.947 for the strongest comparable published study
Where it could be used
Because it needs very little memory and processing power, the model is suited to intrusion detection on small devices in smart homes, factories, industrial facilities and smart cities – places where heavy security software cannot run. The technology is still at the research and development stage and has not yet been deployed commercially.
Source: Gulf News.
