Calculating the speed based on adjacent ADS-B data points will not work. The data is just not precise enough to do this. Suppose that second data point was actually received at 23:11:00.999. If it had been received a thousandth of a second later then it would be timestamped a second later and the first delta t would have been 4 seconds instead of 3 and the second would be 3 seconds instead of 4. That would change the caclulated speeds by about 1/3, which is a couple of hundred knots.
This is true, as I have found out - we can't get valid data between individual ADS-B data points. But if we had a hundred or so of them we should be able to deduce the effective winds aloft component by comparing reported TAS vs calculated groundspeed.
I suppose the data exists in the FR24 database, but that would be a lot of work for a pretty small return of information, unless it's already automated in FR-24's software.
A similar but smaller problem comes from the precision in the coordinates. If the delta in the latitude is .004 in one 3 second interval and .005 in the next, that does not mean the speed changed by 25%.
The difference between 4.091 s 110.345 e and 4.091 s 110.344 e is only about 400 feet or .07 nm and when (in this case) measured from the last data point it only makes a 209 foot difference in distance traveled (4,700ft vs 4,909ft) or about .5%.
Edit:
As Phalanger points out, this has to depend on an assumption of constant altitude.