<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Aug 28, 2014 at 6:17 PM, David Leon Gil <span dir="ltr"><<a href="mailto:coruus@gmail.com" target="_blank">coruus@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class=""><br>
</div>The issue is that usernames are extremely guessable. I think that<br>
Joseph Bonneau had some stats on this in his thesis.<br></blockquote><div><br></div><div>Unfortunately I don't (I have stats on my thesis on the distribution of *human* names, not email addresses). Email addresses are unique, unlike passwords or names, so it's a slightly different statistical problem from what I've worked extensively on the theory side of. Doesn't really matter-bottom line is if you can try a few million guesses for plausible email addresses, you'll probably harvest a large number regardless of the hashing regime involved.</div>
<div><br></div><div>Using a strengthened hash is of course preferable to nothing, as Tony points out there are other ways to harvest email addresses and we can claim this is no worse.</div><div><br></div><div>Still, we're basically having the DNSSEC zone-walking debate again here. Hashing helps in both cases but we are slightly changing the privacy aspects of email addresses/subdomains with a new security feature. Maybe people shouldn't have relied on either thing being private to begin with, but this new proposal does change things and that needs to be carefully thought about.</div>
</div></div></div>