matrix.org/static/jira/browse/SPEC-378

50 lines
1.4 KiB
Plaintext

---
summary: Should m.login.password require a hashed password rather than a plaintext password?
---
created: 2016-03-30 18:21:09.0
creator: neb
description: |-
Submitted by @matthew:matrix.org
To stop leaking human-intelligible passwords to synapse
id: '12604'
key: SPEC-378
number: '378'
priority: '3'
project: '10001'
reporter: neb
resolution: '3'
resolutiondate: 2016-03-30 18:51:55.0
status: '5'
type: '1'
updated: 2016-03-30 18:51:55.0
votes: '0'
watches: '2'
workflowId: '12704'
---
actions:
- author: matthew
body: |-
Answer: probably not, based on:
<eternaleye>
I do have strong thoughts, specifically "please, for the love of $deity, don't"
It's unambiguously worse in every way
It complicates the spec, complicates clients, provides no security benefit whatsoever, and constitutes rolling your own crypto.
In fact, it weakens security, because now if either the prehash or the hash used to store it in the DB are collision-vulnerable, then some types of attacker have an easier time.
This is very close to SPEC-346 so closing as a dup thereof
created: 2016-03-30 18:51:47.0
id: '12800'
issue: '12604'
type: comment
updateauthor: matthew
updated: 2016-03-30 18:51:47.0
- author: matthew
body: dup SPEC-346
created: 2016-03-30 18:51:55.0
id: '12801'
issue: '12604'
type: comment
updateauthor: matthew
updated: 2016-03-30 18:51:55.0