We introduce and solve the problem of Byzantine fault tolerant distributedquickest change detection in both continuous and discrete time setups. In thisproblem, multiple sensors sequentially observe random signals from theenvironment and send their observations to a control center that will determinewhether there is a change in the statistical behavior of the observations. Weassume that the signals are independent and identically distributed acrosssensors. An unknown subset of sensors are compromised and will send arbitrarilymodified and even artificially generated signals to the control center. It isshown that the performance of the the so-called CUSUM statistic, which isoptimal when all sensors are honest, will be significantly degraded in thepresence of even a single dishonest sensor. In particular, instead of in alogarithmically the detection delay grows linearly with the average run length(ARL) to false alarm. To mitigate such a performance degradation, we propose afully distributed low complexity detection scheme. We show that the proposedscheme can recover the log scaling. We also propose a centralized group-wisescheme that can further reduce the detection delay.
展开▼
