I don't know the answer. But let's discuss the problem a bit:

Having three signals A, B and C and wanting them to be inverted individually means to have a system of three degrees of freedom.

But I guess, that the truth table is "rich" enough to solve the problem by letting go one degree of freedom. Means, if we want to decide the status of one signal, it should be enough to ask, what the two others do (without differentiating between them!) in correlation to what all the three do (again not differentiating between them!). That would reduce the number of degrees of freedom from three to two, which at least theoretically could be solved by the help of only two inverters.

Reducing the degrees of freedom seems the only possible way for me, because it's impossible to directly invert an input by using only ANDs and ORs.

Kai