Re: Postgresql -- initial impressions and comments
От | Tycho Fruru |
---|---|
Тема | Re: Postgresql -- initial impressions and comments |
Дата | |
Msg-id | 1038958471.1254.6.camel@bozo.fruru.com обсуждение исходный текст |
Ответ на | Re: Postgresql -- initial impressions and comments (Bruce Momjian <pgman@candle.pha.pa.us>) |
Ответы |
Re: Postgresql -- initial impressions and comments
|
Список | pgsql-general |
On Tue, 2002-12-03 at 23:50, Bruce Momjian wrote: > Scott Lamb wrote: > > That's disconcerting to me because I think it defeats the point of > > sending MD5 signatures on the wire - avoiding replay attacks. If it's > > stored in MD5 format on the server, it can't request it with a different > > salt every time (how would it compare them?), so you can just replay the > > MD5 transmission. > > > > The other way, though, a compromise of the database would mean a > > compromise of all the passwords. > > > > So it definitely would be helpful to have an answer to your question in > > with the description of the authentication types, so you could choose > > intelligently based on what you consider to be more likely risks. > > 7.3 stores encrypted MD5 passowords in database (7.2 it is optional). > We send random salt to client and client double-MD5 encrypts, so > playback will not work --- best of both worlds. So, if I understand it correctly : - on the wire : no cleartext passwords, only doubly hashed + salted passwords -> no replay possible (watch out for session hijacking though) nor password sniffing - in the database : no cleartext passwords are stored, but access to the md5 hashed passwords is sufficient to get access to the database - without really knowing the user's password - by using a modified client. Is this correct ? cheers Tycho
В списке pgsql-general по дате отправления: