We've already translated many dead languages that are no longer spoken. I don't see how translated one that's actually being used could possibly be harder.
Translating a dead language that is no longer spoken is incredibly difficult, and generally requires some starting point. In the case of hieroglyphics this was the Rosetta stone, and also the discovery that Coptic was distantly related to the language of ancient Egypt. With the combination of a similar spoken language, and a passage of text translated into a known language, the task became (relatively) easy. There are plenty of dead languages that do not have these starting points, and have not been translated to this day.
With a still spoken language the task isn't quite so difficult, since the point and name an object approach can get things started, and you presumably have some cooperation from the speaker of the other language. Something along these lines must have occurred when Europeans first arrived in America, since presumably the languages would have had practically nothing in common.
In the case of aliens though, they might well use a different means of communication than speech, which would cause further problems. Translating a language in an unfamiliar form is always going to be harder since you have fewer starting points to work from. A big problem with hieroglyphs for a while was no one was certain in which direction they were supposed to be read (in practice it varies). A more drastic case is the Incan
quipus, which are still mostly untranslated.
I've never been certain about the use of maths as a universal language. For a start an alien culture would probably use a different base than 10, and in any case once you've got the prime numbers, or pi from them, what then? You could send and recieve these in a fairly comprehensible form as a string of bleeps, but it tells you nothing about their language. Maths allows you to talk about maths, but it fairly useless for anything else.