Spam has always been a problem with Amazon’s email service (SES). They make an effort to filter the outgoing missives transmitted by their customers, but it’s not perfect. And Amazon is no respecter of laws outside the good ‘ol US of A, where the right to free speech is a license to spam any kind of junk you like; whether the recipient asked for it or not.
Here’s a case in point:
Received: from a8-55.smtp-out.amazonses.com (a8-55.smtp-out.amazonses.com [184.108.40.206]) by xxx.xxx.xxx.uk (8.14.4/8.14.4) with ESMTP id t5NHpefn075543 for <email@example.com>; Tue, 23 Jun 2015 18:51:40 +0100 (BST) (envelope-from firstname.lastname@example.org) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/simple; s=ug7nbtf4gccmlpwj322ax3p6ow6yfsug; d=amazonses.com; t=1435081898; h=From:Date:To:MIME-Version:Message-ID:Reply-to:Subject:Content-Type:Feedback-ID; bh=jCdtb+gUf4FAvUudtcIKxlX0IOnQHEd/YxIGxHXLcQ4=; b=cNIs7cNe5LzyxYvGWw/LdIeA7epknAFAoeQYjiyf9b5mTKRYLAW9KLvUTSGtlsr7 WWy52wd3Tz9o9vQryvK/Q5l5okAFxgZCZa5uSbXMor7sa/1dU02kwjCyACnb7viR1np BlEytfbGEBUlAfBBrrJueagmdzwa+IXNZsBo4w2Y= DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/simple; s=lfgclj2zbjygv5i5rirpal2v2zj3dquy; d=uebaps.com; t=1435081898; h=From:Date:To:MIME-Version:Message-ID:Reply-to:Subject:Content-Type; bh=jCdtb+gUf4FAvUudtcIKxlX0IOnQHEd/YxIGxHXLcQ4=; b=bZZSEICBkHU8HkdFtiYg9fp+qxzmxJlfNj6UclS3B4dtaKBMTf1oSCSQR5jm0XXE 0JxmIdNWKsgumLUcf8XnZGZFVfwe2f7cVOCiA1EcHX7oHn0weHQjoce+nxwVClgCQYz m0OlXn/YvNBE1MwSvpQR3PfoSCyTVQQpBWjgD8dQ= From: Ray-Ban Sale <email@example.com> Date: Tue, 23 Jun 2015 17:51:38 +0000 To: "firstname.lastname@example.org" <email@example.com> X-MessageID: OXx8fHwxMzY3MXx8fHxmcmFuazJAZmpsLmNvLnVrfHx8fDEwfHx8fDF8fHx8MA%3D%3D MIME-Version: 1.0 Message-ID: <firstname.lastname@example.org> X-Priority: 3 Reply-to: Ray-Ban Sale <email@example.com> Subject: Spambait: Keep Calm and Get 80% Off Ray-Ban! Content-Type: multipart/alternative; boundary="b1_b18fea4f74280e521923210f4d5c61eb" X-SES-Outgoing: 2015.06.23-220.127.116.11 Feedback-ID: 1.us-east-1.E00ipiLUCdDBKP1kTeYjtCc2E2c3DbfGjCtoi1emL2E=:AmazonSES --b1_b18fea4f74280e521923210f4d5c61eb Content-Type: text/plain; charset = "utf-8" Content-Transfer-Encoding: base64 SGksRnJhbmsgTGVvbmhhcmR0OiAjUl9Ub3BfVGl0bGUjLg0KQm9ybiBmcm9tIGEgbWVzaCBiZXR3 ZWVuIHR3byBvZiBSYXktQmFuJ3MgbW9zdCBpY29uaWMgYW5kIHBvcHVsYXIgc3VuZ2xhc3NlcyAt IHRoZSBDbHVibWFzdGVyIGFuZCBXYXlmYXJlciAtIFJheS1CYW5DbHVibWFzdGVyIE92ZXJzaXpl
As you can see (if you’re used to reading email headers), this looks very legitimate – send from a correctly configured server. However. these characters are as guilty has hell. The email body, once decoded, claims that the spambait email address belonged to a past customer of theirs, and was used for placing an order (in the USA). This is, of course, physically impossible.
If this had been sent in Europe they’d have been breaking the local law that implemented the EU Privacy and Electronic Communications Directive, 2002. But they’re sending it from the USA. Other text in the email suggests it’s not from an English-speaking country (not even the USA), and it’s probably a scam. But Amazon doesn’t t seem to mind – they don’t even have an abuse reporting system for ISPs plagued by this stuff.
It’s tempting to simply block all Amazon SES IP addresses, but this will cause collateral damage. Spam filtering isn’t likely to detect it any other way, as the sending server is set up correctly, with SPF records and so on, so the Bayesian filter in a spam classifier will be over-ruled. However, this correctness can be used against it…
Let’s be clear here – it’s easy enough to block the whole of SES. You can get its address range just by looking at it’s SPF records:
%nslookup > set type=TXT > amazonses.com Server: 127.0.0.1 Address: 127.0.0.1#53 amazonses.com text = "v=spf1 ip4:18.104.22.168/22 ip4:22.214.171.124/22 ip4:126.96.36.199/18 -all"
I suspect this may cover more than SES, but SES is certainly covered by it. However, blocking it will, as I mentioned earlier, block some innocent stuff that you do want. This is a job for Spamassassin.
I’m experimenting by adding the following to SA’s local.cf file:
header AMAZON_SES Received =~ /amazonses.com/ score AMAZON_SES 3.5 describe AMAZON_SES Sent from Amazon SES - often used by spammers
The the appropriate score to weight it by is an interesting question. By default good SPF records are ignored anyway; if they were not then it would obviously be a good idea to negate a positive score here. So I’ve picked 3.5 as this matches a clear Bayesian score rather than for any good statistical reason. Check back later to see how well it works.