<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#ffffff">
28.04.2010 15:25, <a class="moz-txt-link-abbreviated" href="mailto:cyrus-sasl-request@lists.andrew.cmu.edu">cyrus-sasl-request@lists.andrew.cmu.edu</a> написав(ла):
<blockquote
cite="mid:mailman.173883.1272482711.5312.cyrus-sasl@lists.andrew.cmu.edu"
type="cite">
<pre wrap="">This is another problem, but server can't encrypt messages because it
should have access to a symetric or assymetric keys saved on the
server. Then you back to the problema, the hacker may access the key
and decrypt messages (this chicken and egg problem).
</pre>
</blockquote>
<div align="justify">My proposal addresses this problem. I fear, you
have not read it -- despite two opportunities already -- before rushing
to the reply-button. Allow me to afford you one more explanation.<br>
</div>
<br>
<div align="justify">The new messages arriving to the server will
remain unencrypted -- no worse off than they are now -- until the user
logs in (and provides the key).<br>
</div>
<br>
Again. Under my proposal, the server does have the key to a user's
messages, but <i>only while the user is logged in</i>. <u>Older
messages of disconnected users are not readable even to the server's
root</u>.<br>
<br>
My proposal does not solve the problem completely, but it does reduce
the damage. This is useful.<br>
<blockquote
cite="mid:mailman.173883.1272482711.5312.cyrus-sasl@lists.andrew.cmu.edu"
type="cite">
<blockquote type="cite" style="color: rgb(0, 0, 0);">
<pre wrap=""><span class="moz-txt-citetags">> </span>The proposed method uses each user's own password to encrypt their mails --
<span class="moz-txt-citetags">> </span>only the mailboxes of the currently-connected users would be exposed to a
<span class="moz-txt-citetags">> </span>hacker (or coercer).</pre>
</blockquote>
<pre wrap="">If the hacker owned the server he can
- use "tcpdump -s 0 -A | grep --line-buffered -e LOGIN -e USER -e
PASS" to get password in next user authentication.
- read TLS private key file and look traffic with tcpdump.
- read TLS private key from memory.
- switch imapd daemon to a version that save user/password on a file.
</pre>
</blockquote>
<div align="justify">This would only give the hacker ability to access
e-mails of people currently connecting to the server, while the exploit
is ongoing. My plan -- for the third time -- aims to protect mailboxes
of those, not currently connected. I believe, this could be valuable in
a substantial number of installs.<br>
</div>
<blockquote
cite="mid:mailman.173883.1272482711.5312.cyrus-sasl@lists.andrew.cmu.edu"
type="cite">
<pre wrap="">Server should't encrypt data. Root can do anything.
</pre>
</blockquote>
<div align="justify">Both statements are wrong (as all
generalizations)... I demonstrate, how the server can do encryption
usefully -- so that even root can not decrypt it, until the user logs
in to check their e-mail.<br>
</div>
<br>
<div align="justify">If a break-in happens, while a I'm on vacation, my
old e-mails weren't exposed. That's as useful as being able to lock my
house, while I'm away, even if I have to open it up upon returning...<br>
</div>
<blockquote
cite="mid:mailman.173883.1272482711.5312.cyrus-sasl@lists.andrew.cmu.edu"
type="cite">
<pre wrap="">Server should't encrypt data.
</pre>
</blockquote>
You are now contradicting your own earlier advice (to use encrypted
filesystem)!<br>
<br>
<div align="justify">I must say this explicitly, Reinaldo, that you are
coming off as an annoyed, impolite, and discourteous individual and
this is my last e-mail to you, unless your response shows better
manners. I don't want to emulate your tone <i>again</i>...<br>
<br>
This need not be -- and is not -- about anybody's ego, you know...<br>
<br>
Yours,<br>
</div>
<blockquote>-mi<br>
</blockquote>
</body>
</html>