Why you will like Canary Tokens

Ever heard of Canary tokens? These are the digital equivalent of the canaries that were used to detect gas in coal mines. The coal mine canary dies of the gas before miners do and give the miners time to save themselves. Similarly, the canary tokens you plant in your systems can give you a heads up that a hacker is snooping around. This is typically in the reconnaissance fase and you might even have a chance to prevent lateral movement or exfiltration.

As with the real life canary, you can have a really low false positive rate. Given the huge amounts of false positives leading to "alert fatigue" in security teams this is really welcome. The trick of course is that you set up your canaries in such a way that no regular user is likely to trigger them. If you have a word document named "vulnerability report" in your home directory, nobody other than you should open it. And you of course no that opening this fake document sets of a trap and you leave it alone. If you're using the same trick to find snoopers in group directories, you better make sure that everyone in the group knows about the trap. If you're exclusively looking for evil hackers and not for internal leaks you could also consider having the directory hidden from normal users.

There are lots of different types of canaries available: fake documents, URL, trigger, DNS triggers SQL Server triggers, AWS triggers and more. You can easily expand on these with any action that you can translate in one of the existing triggers. An example is the SQL Server trigger, where a certain SQL statement will cause the database server to do a DNS lookup that in turn triggers the trap.

You can easily setup a canary token on canarytokens.org. The server running there seems to be running a bit behind on the github project, and we're thinking of running an instance at ShadowTrackr.com or another domain that might be a bit less of a giveway for those that monitor dns traffic. You can use ShadowTrackr to turn the webhooks that canarytokens call into push messages on your phone and log them to have an audit trail for post mortems. See this use case for an example.

More scanner nodes and new locations

Things are going well and the total number of assets under monitoring is ever increasing. That means that scanning capacity had to increase too. Unfortunately, this was lagging a bit. Some host and website scans where consistently running behind schedule.

Adding new scanner nodes is a job that only requires a few minutes per node. It's not fully automated (yet), but not really a major hurdle too. What was a challenge is making sure the central database and webservers remain snappy and don't grind to a halt.

The good news: It has been fixed. Besides the extra nodes we've also added extra datacenters: Mumbai (India) and Toronto (Canada) are now scanner node locations too. The extra capacity should kill the backlog in a few days. If not we'll be adding extra nodes again later this week.

How to monitor keywords

Keyword monitoring needed improvement. Recently I was confronted with a possible dataleak and wanted to monitor a few specific names and phrases in the news. The way it worked before was that the bots checked for the word with a space before or after it in the text on newsites and copy-pastesites:

        if " "+keyword in text or keyword+" " in text:
    

While it does a reasonable job of not matching things like "blakeywordbla", it had way to much false positives. I needed more finetuning and ended up implementing literal matches, multiple keyword matches and negative keywords.

Literal matches: "the keyword"

It works just like people are used to in Google searches. The exact words need to be present in the order specified. It triggers on "This is the keyword you are looking for", but not on "this keyword is not the one you are looking for".

Multiple matches: the keyword

Both "the" and "keyword" need to be present in the text in order to trigger, but the order is not important. They can easily be sentences apart, like in: "This keyword is what you want. The other words are ignored."

Negative matches: keyword -the

Again like in Google searches, any text with the word "keyword" in which "the" does not appear at all is a match. For example: "The keyword is not enough" doesn't match, and "Keyword matches are very useful for finding leaked data" does.

Of course you can mix all of the above. Do note that all keyword matches are case insensitive, and that news articles from the newsfeeds that we monitor trigger both on keyword matches in the headline and the text. Datadumpsites often don't have a title and matches are on keywords in the dump itself. Here are some keyword combinations to get your creativity started:

shadowtrackr leak -water
@shadowtrackr.com
shadowtrackr password
"Tracking your online footprint"

You can add them under Assets in the sidemenu . Happy keyword hunting! And don't forget to set push notifications to get a heads up on those really bad days.