The numbers may indeed be random, and they may be from a physical process, but are they actually quantum random? As the article indicates, it is indeed true that there is no method of determining whether a string of numbers that looks random is actually random. You can tell if it's not random through the use of various statistical tests, but there may always be a test that hasn't been invented yet that could reveal an underlying pattern.
However, distinguishing between quantum randomness and "other" randomness is quite easy actually. You just ask: did the randomness come from the collapse of a quantum state into an eigenstate of the observable being measured? (Now, if it's already in an eigenstate, it will collapse to the same state, so you have to prepare it in an eigenstate of an incompatible observable before measuring a new observable.) The particular new eigenstate that it collapses into will be -- according to the postulates of QM -- truly random.
So it seems to me that with this "QM random" source from the cell phone sensor, the scientists should be analyzing the "purity" of the signals they are using to generate their numbers. All kinds of factors could make the results non-QM random: electrical noise, different kinds of light sources, processing delays, etc. There's a lot of things to consider here.
(In fact, if you really wanted to be sure that it was QM random, you could create an entanglement [which is a QM-only phenomenon] and then see if the Bell correlations exist between the two entangled states.)
Couldn't you just take a picture with the same camera and then hash it? The probability of getting the exact same set of pixels is small and it's certainly not predictable.
I wondered the same thing. Noise in the picture should also make it extremely hard to predict. For cryptographic use one would only need 256 bits of entropy to seed a CPRNG, and I can't see how this amount of entropy wouldn't be present in a normal photo.
