Surveillance Detection Scout - Your Lookout on Autopilot

Conference:  Defcon 27



The presentation discusses the development of an experimental software project that uses a Tesla car's cameras to capture footage and process it for various purposes.
  • The software project involves setting up partitions on a Xavier device to allocate storage space for captured footage
  • A crontab is used to run a check copy script every 30 seconds to unmount, wait, and mount a USB flash drive that stores the footage
  • The footage is processed using ffmpeg to remove errors and convert it into frames for inference
  • Third-party services such as Mapbox and OpenStreetMap are used to enrich the geocode data and provide reverse geocoding
  • Drives are created programmatically to track the car's path and activities
The speaker mentions dealing with corruption issues in the video files and using ffmpeg to fix them. They also note that the project is experimental and should be used cautiously to avoid losing important footage.


Surveillance detection routes are a daily occurrence for clandestine operatives and agents all over the world. These mentally taxing counter-surveillance measures often mean the difference between life and death. Surveillance Detection Scout hopes to ease that burden. Scout currently supports Tesla Models S, 3 and X, running license plate recognition on 3 camera feeds to alert you in real time if you're being followed. When you park, Scout remains vigilant, implementing familiar face detection as well. By combining timestamped vehicle location data & video, computer vision and an intuitive web interface, it becomes apparent that Scout has just as many offensive as defensive applications. Over time, SDS captures and reports on observed patterns of life, allowing you to quickly gain an overview of your surroundings (or your target) with minimal effort. Whether you're conducting or evading surveillance, Scout has got your 6.



Post a comment

Related work