Paper - Detecting Wireless Spy Cameras Via Stimulating and Probing

  • Metadata:
    • author: Tian Liu, Ziyu Liu, Jun Huang, Rui Tan, Zhen Tan
    • title: Detecting Wireless Spy Cameras Via Stimulating and Probing
    • year: 20182018

    • url: Weblink
  • Essay:
    • With wireless security cameras becoming cheaper and smaller, the bar for privacy invasion is lowered, upping the demand for detection tools. As traditional methods of detecting these devices (e.g., RF-based probing) are cumbersome, new ways of combating spy cameras are being researched.
    • The system proposed by this paper aims to detect these cameras by executing two steps: They first change the lighting conditions in a room, making the captured footage suddenly spike in size due to the compression/encoding algorithms commonly used by these devices. They then analyze the network traffic on the mac-layer to notice devices with changes in their transmission rate (indicating a high probability of spycams).
    • They deploy their approach in two forms: Blink, a mobile app that requires some manual operation, and Flicker, a cheap hardware circuit that can automatically detect cameras with differing modes of transmission. By working around issues like uneven packet flows and ambient lighting, they manage to achieve a detection accuracy of 90% and above for both of the systems.
    • I enjoyed how the paper turns the discussion about phones in a privacy context around. Rather than the smartphone being the perpetrator that invades our privacy, it is used as a tool to keep up said privacy.
    • The simplicity of their approaches also really resonated with me. While the work on fixing the more fine-tuned challenges in packet analysis is undoubtedly on the more sophisticated side, the basic idea is a classic case of putting two concepts together: The behavior of video encoding under changing light conditions and the concept of network analysis. Research like this is what motivates me to look for similar idea combinations in my surroundings that could produce interesting results when executed (e.g., dating apps + no internet connection).
    • Some points stuck out to me a bit: They mention that building around their systems would substantially increase manufacturing and deployment cost for the spy cameras. Even though some changes would need to be made, equipping these cameras with a little bit of buffer storage (think of the size of MicroSD cards) and changing up the network protocol could easily fool the presented approaches, while not increasing the prices significantly. Adding randomization into the transmission (delays, quality variation) would be easy to implement and could make changes produced by the stimuli much harder to detect.
    • What bewildered me the most is the fact that they never mentioned the phone's flashlight as a possible light source. Automatically controlling the light with the phone could simplify the usage of Blink by a lot and reduce the needed hardware for Flicker (even though the light would be human-visible in this case). I suspect there are good reasons that would prevent the flashlight from being usable, but a brief discussion of these would have been a welcome addition to the paper.
    • However, as this is the first execution of an idea, these issues can be overlooked and I am interested to see how their concept will be developed further.