Probably all of the above mentioned are good sources. I sometimes got cell lines directly from the original scientists (who often don't want to go through the work of supplying the whole world with their cell lines... so they have the right to ask for a "handling fee" and a Material Transfer Agreement (MTA), which may take some time to get it approved (and heavily changed) by the lawers of the university... Sometimes there is no written agreement necessary, but I once needed about 5 years to get a published cell line (some people told my I was the only one who ever got a cell line from there...). Many times you may easily get cell lines from your collaborators, or just neighboring colleagues. This may be the fastest and easiest way - and could be totally OK, but you asked for authenticity. There is always a risk you get a wrong cell line - sometimes cells really are not of the origin they were described - I try to write a paper with wonderful results with a cell line, which is not really what it was supposed to be, but showed most of the characteristics of a specialized cell type, but (happily) not all signs - so this was good for the functional experiments.
Another problem is: many cell lines and labs have problems with mycoplasma. Some people can get in Science and Cell, but it some institutions have clear guidelines (for example, a necessity of testing all cell lines for an animal experiment for such infections...).
About the danger of getting the line not from the cell bank - my colleague from the lab have made an analysis of three U937 cell lines, one from ATCC, one from our lab and one from another lab showing the phenotypic differences between the cell lines obtained from different sources (the one in our lab was identical to the one from ATCC, but the one from another lab was very different but still U937).
Cytokine. 2001 Mar 21;13(6):365-70. Identification of two U937 cell sublines exhibiting different patterns of response to tumour necrosis factor. Kaszubowska L, Engelmann H, Gotartowska M, Iliszko M, Bigda J.
We got a very nice model from this but it was not the first aim of this work.
So the risk is even higher - You can get not only the wrong cell line or cell line infected with mycoplasma as Robert Eibl has mentioned, but even worse: a right cell line slightly mutated.
So be very, very careful with obtaining the cell lines not from the cell bank which will give You certificate of the authenticity for the cell line. Especially, that such a certificate (for human cell lines) is a condition for publication in many journals lately.