This past week was truly one for the blooper reel. A public cloud service provider let the great unwashed into the address ranges published as safe mailers via their SPF records, with hilarious if rather predictable results. Next up, we find an intensive advertising campaign for spamware aimed at our imaginary friends. And the password guessing aimed at an ever-expanding dictionary of non-existing users continues.
To the rest of the world, bsdly.net is known variously as a honeypot, a source of various kinds of blocklists or a frequent target of domain joejobs that contribute to the ever-expanding list of imaginary friends also know as spamtraps.
To me and a very small set of other people, it's home on the net, providing a set of services we need fairly painlessly on an OpenBSD platform that rarely requires much work besides the odd pkg_add -u -D snap followed by sysupgrade -s (yes, we jump from snapshot to snapshot on this one).
Then the past week served us with three separate events that, while actually harmless to our side, together serve to show that a certain subset of humans would perhaps be better diverted to activities that do not involve computers.
Note: This piece is also available, with nicer formatting but unfortunately also Blogger's trackers, here.
The first event started on Tuesday. While looking for something else entirely in my mail server logs, I noticed an unusually high number of delivery attempts to what definitely looked like spamtraps reaching the actual mail server. The earliest entry was from that morning:
2022-03-29 06:28:31 H=ministeriodesanidad34.yomevacunoseguro.com [18.104.22.168] X=TLS1.2:ADH-AES256-GCM-SHA384:256 CV=no F=<firstname.lastname@example.org> rejected RCPT <email@example.com>: Unknown user
There were several thousand entries of that type (full log here, extracted source IP addresses here). My initial impulse was of course to check the logs to see how they got past spamd in the first place. Oddly, I found no trace of any activity involving spamd and a random sampling of those IP addresses. That would in turn indicate that they had been pass-listed, most likely by being included in the permanent nospamd pass list, which we generate here mainly based on the published SPF records of domains we need to communicate with.
Again taking a random subset of the extracted IP addresses and using whois identified the IP address range owner as a large cloud services provider that among other things also provide hosted or hybrid on premise plus hosted email service. This in turn means that domains that use those services also include large segments of that provider's IP address ranges in their SPF records. Not quite knowing what to do I tweeted,
Grumble.— Peter N. M. Hansteen (@pitrh) March 29, 2022
Spam campaigns sending from ranges in our #nospamd tables (and hitting actual SMTP service) due to being in a major #cloud operator's #SPF records, sending to a large chunk of our #spamtraps.
Still developing, blogworthy, Y or N?
In addition to tweeting and looking for feedback (which was not huge but dominated by Y answers), I notified the relevant abuse@ address by mail, including links to the log data and the IP addresses.
I also tweaked the log reader hinted at in this earlier piece so any attempt at delivering mail from that domain in the future will put the sending IP address safely away both in the spamd blocklist and in the safety of a table that is subject to block drop quick from for six weeks after the activity stops, and exported to downloadable blocklists as described in the article I referenced earlier.
The abuse@ handlers at the company I am not naming explicitly here (yes, there's a unix command you can use to find out who if it matters to you) were quite responsive and said that the activity seemed to be coming from their public cloud section, and yes they were forwarding my data to their internal CERT. As a followup I suggested to them that using our slowly expanding list of spamtraps in their outbound filtering might be a good idea if they intend to offer SMTP-for-hire in the future.
What seems to have happened is that the miscreants here set up using a range of the provider's services including domain registration, DNS hosting, and judging from the consistent use of root@ as the sender addess, set up some number of Linux virtual machines to do the spamming.
Before the activity stopped later in the week, we identified two more campaigns that fit the pattern. The data can be found here: log entries and IP addresses for the second wave, log entries and IP addresses for the third. Each of the campaigns appear to have stopped shortly after their domains were de-registered. I never saw the contents of the messages since not a single one appears to have inboxed here.
The episode has a few teachable items: First, that some subset of our list of spamtraps is indeed incorporated in the address lists used by gullible spammers and their customers, and second, that if you run a public cloud service, you need to pay attention to what your customers do and be wary of letting them use IP address ranges that have been announced as being really safe to accept mail from.
I notified the cloud provider that I would be writing an article about the events and asked them for any and all useful input they could provide. No such information has surfaced by the time of writing. If any useful information turns up from them after publication, I will of course update this piece accordingly.
While the public cloud spammers thing was developing, I noticed another campaign that was actively targeting our spamtraps.
Sent from as far as I could tell only three IP addresses and with a total of 58 different subject lines all about spamming tools. It is possible that the campaigns did not target our spamtraps exclusively in our domains, but the log archive (1.5M compressed but expands to 40M raw) serves as testament that our imaginary friends are definitely targeted by some subset of the online marketing community, and they are pouring resources into doing that, one byte per second.
And once again, I have not seen the actual contents of the messages beyond what turns up in the logs after greytrapping kicks in. Not one of those messages found its way to an actual mailbox here.
As noted in a previous article, SSH password guessing activity went up significantly in the days leading up to the Russian invasion of Ukraine i February.
In addition to the data referenced directly in the article, the archived logs and the summaries of numbers of attempts per day (March and April so far) the number of attempts per day are consistently on a higher level than before the Ukraine war started, with a new higher intensity episode ongoing as I type.
One interesting feature of the password guessing attempts during the last few days is that they feature a much larger number of new user names attempted than usual. This means that the list of spamtraps here is now growing at the highest rate since the episode involving what was likely one or more phishing campaigns targeting Chinese users during early 2019 as mentioned in my 2019 summed up piece published at the end of that year.
If you have further data on these or similar incidents that you are able to share or if you want to look further into these and similar incidents, please let me know.
If you find any errors in the material I publish or disagree with my sentiments, or if you find this article interesting, useful or annoying, please let me know, either in comments or via email.