Early detection of clandestine tunnels beneath critical infrastructure such as oil and gas pipelines represents a major technical challenge. Traditionally, monitoring methods only identify damage once the pipeline has been compromised, which can lead to fuel theft, sabotage, or even environmental disasters. Ground-penetrating radar (GPR) offers a non-invasive alternative for visualizing the subsurface, but manual interpretation of radargrams is impractical for continuous monitoring along extensive corridors. Furthermore, supervised approaches require examples of tunnels that are scarce in practice. In this context, a fully unsupervised detection system, based on depth-constrained scoring, has demonstrated exceptional performance: it achieves an AUC of 0.994 and an F1 score of 0.975 without needing labels for training, scoring, or threshold calibration. The key lies in a denoising convolutional autoencoder that learns the structure of normal terrain and then identifies anomalies through reconstruction error. By limiting the analysis to the depth band where tunnels are physically possible, the technique drastically reduces false alarms and improves accuracy.
This approach illustrates how artificial intelligence and advanced signal processing techniques can solve critical security problems in infrastructure. At Q2BSTUDIO, as a software and technology development company, we apply similar principles to create custom applications that integrate AI models, whether for anomaly detection, process optimization, or predictive analysis. Our experience in AI for businesses allows us to design solutions ranging from computer vision systems to real-time monitoring platforms. Additionally, we combine these developments with AWS and Azure cloud services to ensure scalability, availability, and data security. We implement AI agents that automate inspection tasks and alert on anomalous behaviors, reducing reliance on human supervision and improving operational efficiency.
Cybersecurity also plays a fundamental role in these environments: protecting both data acquisition systems and AI models from attacks is a priority. Therefore, we offer business intelligence services with Power BI to visualize performance metrics and detect risk patterns. Likewise, the custom software we develop integrates with cloud platforms, enabling centralized sensor management, data pipeline orchestration, and continuous model updates without disrupting operations. The combination of these technologies not only improves critical infrastructure security but also adds value in sectors such as mining, construction, or geotechnical exploration, where detecting tunnels or underground voids is essential.
The presented case demonstrates that, with intelligent design of the scoring architecture and leveraging physical domain knowledge, it is possible to achieve superior results even with minimal or no labeled data. At Q2BSTUDIO, we bring this philosophy to our projects: each solution is tailored to the client's specific context, employing machine learning and signal processing techniques that maximize accuracy and minimize false positives. If your organization faces similar challenges in unsupervised detection, asset monitoring, or geospatial data analysis, we can help you build a robust and scalable platform, integrating AWS and Azure cloud services, AI models, and business intelligence tools to transform data into strategic decisions.
