If your steps in the proof are all reversible (so "<=>" type steps), then to prove an inequality, it is logically valid to start with the given to-prove inequality and simplify it down using such reversible steps to something that you know is clearly true (e.g. (x-y)^2 >= 0), and then...