I'm aware it's neither differential nor >1GHz, and well, it's only available in German, but the most elaborate take on a DIY active probe I stumbled over in the last years was this one:
http://welecw2000a.sourceforge.net/docs/Hardware/Aktiver_Tastkopf_mit_OPA659.pdfThe probe uses an OPA659 with a trimmable capacitive divider. The main developer and author of this PDF "branadic" is also active in this forum btw.
Building some of these is one of multiple projects on my "stack".
As a side note: I also bought some of the probes by Alfred Rosenkraenzer (PCB version). When I bought the first one, I was well aware that the bandwidth sounded a bit optimistic. Still, my main intention was to measure LVDS signals in the 20-40MHz range and the probes worked excellently for that. I.e. where the passive 500MHz probes just showed a sine shaped signal, I got a clear rectangular signal with steep edges. I already used them for 5V signals as well though (e.g. 100ns CS pulses) where they are also far superior to 500MHz passive probes. Admittedly, the scoped used was a 600MHz Lecroy so I don't really care if the probes really have a much higher bandwidth or not. And yes, looking at the differential amplitude, the probes show a 20:1 attenuation, I guess the 10:1 value was meant for each input. In the meantime, the description in the eBay offers was also changed as far as I can tell.