| ||
> (How's this for a new keyword?) Simple, self-explanatory, GOOD.
> Now with 26 people, we have 325 pairs, (26 * 25 / 2). > To compute the probability of having a birthday match The easiest way is to think in terms of probability of NOT having any match. The 1st person leaves 364 days open, so the 2nd person has a 364/365 chance of fitting a birthday among those. That leaves 363 days open, so the 3rd person's chance is 363/365. And so on, for P persons the probability of having distinct birthdays is (364/365)*(363/365)*(362/365)*...*(365-P+1)/365, and the probability of a match is 1 minus that. Incidentally, I believe the minimum number of persons to yield a match probability >50% is 23 (not 26) but I just base that on a quick Saturday-evening numbers run so don't bet your fortune on it. > For 20,000 files [...] with CRC-32 ...the chance of a CRC match would be about 4.5%
> However, in the context we are discussing,
As for numbers (from the same source disclaimed above), the chance of a false CRC match surpasses 50% for Now, 77k files sounds like a lot, but if all ZTree users do it over and again, it will happen. And that's without even getting into "intentional" CRC masquerading. There were viruses (virii ?) which would manipulate a few junk bytes to keep the overall CRC from changing for a couple of popular polynomials. Besides, it's also a matter of image and of trust. Did you feel, a few years back, that Intel's floating-point bug would directly threaten you? ...Statistically not. Did you think it's a bug, nevertheless? ...I did.
Liviu | ||
|
Messages in this Thread
- [Discuss] Probabilities and the CRC (915) - John Gruener - Oct 22,2000 at 09:34 [Discuss]
- [Discuss] 23, 302, 77164 (846) - Liviu - Oct 22,2000 at 13:51
- [Discuss] 23, 302, 77164 (1,042) - John Gruener - Nov 02,2000 at 20:38
- [Discuss] 23, 302, 77164 (863) - Liviu - Nov 02,2000 at 21:16
- [Discuss] 23, 302, 77164 (865) - Walter Rassbach - Nov 03,2000 at 09:00
- [Discuss] 23, 302, 77164 (1,042) - John Gruener - Nov 02,2000 at 20:38
- [Discuss] Probabilities and the CRC (1,058) - Walter Rassbach - Oct 22,2000 at 18:31
- [Discuss] Probabilities and the CRC (874) - R.O. - Oct 26,2000 at 00:20
- [Discuss] Probabilities and the CRC (817) - Walt - Oct 26,2000 at 03:38
- [Discuss] Probabilities and the CRC (854) - Walter Rassbach - Oct 26,2000 at 06:28
- [Discuss] Probabilities and the CRC (849) - John Gruener - Nov 02,2000 at 21:09
- [Discuss] Probabilities and the CRC (857) - Walter Rassbach - Nov 03,2000 at 00:12
- [Discuss] Probabilities and the CRC (854) - Walter Rassbach - Nov 03,2000 at 08:56
- [Discuss] How to SPOOF the PKZIP CRC (960) - Walter Rassbach - Oct 23,2000 at 17:28
- [Discuss] How to SPOOF the PKZIP CRC (1,175) - Liviu - Oct 26,2000 at 02:23
- [Discuss] How to SPOOF the PKZIP CRC (919) - Walter Rassbach - Oct 26,2000 at 03:07
- [Discuss] How to SPOOF the PKZIP CRC (1,175) - Liviu - Oct 26,2000 at 02:23
- [Discuss] 23, 302, 77164 (846) - Liviu - Oct 22,2000 at 13:51