Pareto-Optimal Covert Channels in Sensor Data Transmission
Carina Heßeling, Jörg Keller · 2022
If a stream of sensor data values is transmitted as floating-point numbers, a covert channel can be realized by re-coding the numbers into a more compact number format and using the freed bits to transmit a secret message. Yet, re-coding may introduce rounding errors. By applying a uniform description for a large family of number formats, we compute a Pareto front of optimal covert channel bandwidth versus average relative error. Our only assumption is the knowledge about the statistical distribution of the exponent values in the original floating-point numbers. We exemplify such Pareto fronts for several distributions and identify number formats with relevant frequency of appearance.