Secure3D-CV: A scientific framework for integrity assessment of 3D data with an external Python library

<ns3:p> Background Three-dimensional (3D) perception is widely used in robotics, autonomous vehicles and medicine. These systems often operate autonomously and are vulnerable to data tampering or spoofing attacks. Existing work on 2D adversarial detection does not generalise well to the high-dimensional and structured nature of 3D point clouds. Methods We propose Secure3D-CV, a method that builds on OpenCV by adding a library for outlier detection and threat assessment in 3D scenes. The method transforms each scene ( <ns3:italic>S</ns3:italic> ) into a feature vector ( <ns3:italic>x = φ(S)</ns3:italic> ) via geometric, spectral and volumetric descriptors. A statistical anomaly score ( A ( S ) = | x − { x ¯ } | 2 ) measures the deviation of ( <ns3:italic>x</ns3:italic> ) from a benign distribution ({ x ¯ }). A supervised or one-class classifier ( <ns3:italic> f <ns3:sub>θ</ns3:sub> </ns3:italic> ( <ns3:italic>x</ns3:italic> )) estimates ( <ns3:italic>P</ns3:italic> ( <ns3:italic>y</ns3:italic> = 1| <ns3:italic>x</ns3:italic> )) where ( <ns3:italic>y</ns3:italic> ) denotes a safe scene. The final decision score ( <ns3:italic>R</ns3:italic> ( <ns3:italic>S</ns3:italic> )) strategically combines ( <ns3:italic>A</ns3:italic> ( <ns3:italic>S</ns3:italic> )) and ( <ns3:italic> f <ns3:sub>θ</ns3:sub> </ns3:italic> ( <ns3:italic>x</ns3:italic> )) to classify the scene as safe or not safe. Results On two datasets, simulated point clouds and depth maps containing 2,000 samples, and 100 anonymised CT volumes, we show that Secure3D-CV detects injected outliers, occlusions and adversarial perturbations with high true-positive rates (99.3% on synthetic data and 98.7% on CT data) and low false-positive rates (below 1%). We demonstrate that the hybrid ( <ns3:italic>R</ns3:italic> ( <ns3:italic>S</ns3:italic> )) approach provides a significant advantage over purely statistical or machine learning-based methods. Conclusions Secure3D-CV is a lightweight extension to OpenCV that monitors incoming 3D data streams, extracts robust features and computes risk scores to flag unsafe scenes. The library provides Python bindings and can be integrated into robotics or medical-imaging pipelines. This work contributes a principled threat model for 3D data and demonstrates that combining anomaly scoring with a learned classifier fundamentally improves detection performance compared with baseline thresholding and state-of-the-art methods. </ns3:p>