Hey HN,
We are a defense technology team operating out of the Denver/Boulder aerospace corridor. Recently, the security research community highlighted a critical flaw in standard Autonomous Target Tracking (ATT) matrices used by C-UAS drones: The Distance-Pulling Attack (DPA).
If a tracking drone relies heavily on RGB cameras for depth perception and bounding-box generation, a hostile operator can deploy an adversarial umbrella (a "Flytrap"). These feature high-contrast, geometric patterns explicitly engineered to spoof the drone's optical sensors. The drone miscalculates the physical distance, forcing it to continuously close the gap until the operator can capture or destroy the hardware.
To patch this, SkyGuard Defense Systems partnered with InSitu Labs to bypass RGB limitations entirely. We built the Optical Threat Isolation System (O.T.I.S.).
O.T.I.S. is a ROS-native node that strips away optical camouflage. Instead of looking at the picture, it generates a real-time Live Depth Perception (DP) mesh—essentially a structural X-Ray of the battlefield.
// Conceptual execution pipeline
if (optical_variance > DPA_THRESHOLD) {
initiate_override();
deploy_structural_mesh();
target_lock = isolate_geometry(anomaly);
execute_kinetic_intercept(target_lock);
}
When the bounding box detects geometric spoofing, O.T.I.S. overrides the standard feed, isolates the raw structural math of the target, and calculates a zero human-latency kinetic intercept before the drone is pulled into the danger zone.
We built a frontend UI to simulate exactly how O.T.I.S. handles the spatial math during an intercept, including the X-Ray PIP deployment. You can run the live demo below.
We’d love feedback from anyone currently working with ROS navigation stacks or computer vision obstacle avoidance on how you are handling adversarial physical inputs.