Objective:

With the increasing reach of IoT and advanced digital technologies, real-time monitoring of the complex systems with IoT systems has become usual practice. However, due to the connected ecosystem of the sensor networks, such systems are prone to various cyber-attacks. This is an extremely serious situation, leading to an incorrect diagnosis of devices, leading to major operational failures.

Eugenie analyzed data from a secure water treatment facility to detect anomalies such as water leakage or chemical injections. Eugenie’s AI-based insights were needed to differentiate accurately between a genuine anomaly or an artificially induced anomaly, as a result of a cyber-attack. 

Solution: 

Eugenie’s robust digital ecosystem built a digital twin of the distribution process for real-time monitoring of the operational system. A predictive model was developed to detect unusual instances which may indicate a probability of a cyber-attack on a dataset, comprising of an extensive multi-variate time series. 

The technical architecture of Eugenie, comprising of deep learning-based advanced anomaly detection model was highly successful to detect, predict, and diagnose operational anomalies. 

Impact: 

Eugenie’s machine learning algorithms detected anomalies with more than 90% accuracy. This was achieved mainly by using an AI-based framework - Generative adversarial Networks. The data science models applied for the unstructured anomaly detection involved the models Binary Cross Entropy GAN and Least square GAN, with both the models achieving results of 94% and 95% respectively. The accurate insights of AI-driven systems like Eugenie play a vital role in securing industrial systems with prompt decision-making to mitigate risks.