Re: [saag] Ubiquitous Encryption: spam filtering

[Phillip Hallam-Baker <[e-mail]>]

On Tue, Jun 30, 2015 at 2:27 AM, joel jaeggli <[e-mail]> wrote:

> On 6/29/15 8:13 PM, Phillip Hallam-Baker wrote:
> >
> >
> > On Tue, Jun 23, 2015 at 11:19 AM, John Levine <[e-mail]
> > <mailto:[e-mail]>> wrote:
> >
> >     I can't find in the archives whether the ubiquitous encryption
> >     discussion has looked at the knotty issues of spam filtering.
> >
> >     It's a really hard problem -- filtering is essential to keep mail
> >     usable, both due to the sheer volume of the spam and the need to keep
> >     phishing and malware away from recipients.  You can do some filtering
> >     on the envelope, but there's no substitute for looking at the
> contents
> >     of the message.
> >
> >     All of the middlebox issues apply, it's much easier to do the
> >     filtering on a large shared server than at endpoints.  Partly that's
> >     because the endpoints often have limited bandwidth and compute power
> >     (think phones) and partly it's because effective filtering needs to
> >     consult shared frequently updated lists of malware signatures and
> >     malicious urls.
> >
> >
> > I don't think it is actually much of a problem in practice. People just
> > have to be prepared to accept that they probably don't want end-to-end
> > encrypted mail from people they don't know. Once that is accepted, the
> > solutions are fairly straightforward, the publicly visible email
> > encryption key is to the spam filter, after a reply send an end to end
> > key...
> >
> > Main constraint is that you don't want to accept end-to-end encrypted
> > email unless it is signed by someone you know. So the endy mail problem
> > becomes an introduction problem.
>
> The fact that parties that are known to each other have not in general
> been mutually authenticated is a, if not the, significant conduit of
> phishing.


The first last and only reason phishing is possible is that we use
authentication credentials that we expect people to keep in their head,
never write down and only ever give them to people who are trustworthy.

Needless to say, the result has been a fiasco.

If a tenth the amount of whining that goes on complaining about the
occasional failures of the part of the Web security infrastructure that
(mostly) works had gone into avoiding relying on usernames and passwords
thirty years after crack.