Non-dyadic wavelets for detection of some click-fraud attacks

Year to year, the amount of the context and pay per click advertising grows rapidly, and so does the value of its. At the same time, the problem of click fraud is still open. In the paper, we present a novel method of using non-dyadic wavelets for struggle against some kinds of click fraud by attack pattern recognition. Optimal wavelet basis choice is based on the comparative analysis of existing approaches in non-dyadic wavelets construction. We also provide a real-life example to illustrate method.