You do realize in order to perform a double spend you need at least 51% of the network power.
That is not true.
0-confirm tx can be double spent by a mere race attack.
0-confirm tx can be double spent by a Finney attack.
A confirmed tx can be double spent by any significant % of global hashing power although as the number of confirmations rises the probability of success greatly declines. At 6 confirms for example someone with 10% of global hashing power only has a roughly 1 in 10,000 chance of successfully building a longer chain than the rest of the network.