Timeline for Uniform vs discrete Gaussian sampling in Ring learning with errors
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 26, 2023 at 0:33 | comment | added | Lev Knoblock | You mentioned that the sum of two gaussians is a gaussian - isn't the sum of two uniform random variables in a finite field also a uniformly random variable in said field? Or am I missing something? | |
| Jan 2, 2019 at 16:27 | comment | added | Thomas Prest | This is also my understanding. In addition, there are some situations where we still don't know how to remove Gaussians (except at the cost of a huge overhead in performances): trapdoor sampling (introduced in eprint.iacr.org/2007/432), which is used in hash-then-sign schemes, identity-based encryption schemes, etc. It is not directly relevant to the question, but still worth mentioning IMHO. | |
| Dec 31, 2018 at 15:09 | comment | added | TMM | +1 - I suppose if sampling issues were not a problem, everyone would use Gaussians for their optimal trade-off between size and entropy. Alternatives are mostly studied exactly because of these sampling issues. | |
| Dec 30, 2018 at 22:37 | history | edited | Thomas Prest | CC BY-SA 4.0 |
added 154 characters in body
|
| Dec 30, 2018 at 22:31 | history | answered | Thomas Prest | CC BY-SA 4.0 |