Regex for normalizing VERP styled e-mail addresses

[Simon]

Hi all

For better bounce handling, we're using VERP styled from-addresses when sending mails through our postfix. So when splunk parses the mail logs, I have values in the from-field like this:

from=<[email protected]>

Now I'm searching for a regex for search time extraction to remove the VERP id (all after the +).

I tried to use a lookahead ?=, which when matches doesn't get added to the whole mach:

\<(?<realfrom>[a-zA-Z]+(?=\+{1}[a-zA-Z]+)@.*)\>

But this didn't work so far. Any ideas how to get rid of the VERP id?

Thanks Simon

[ziegfried · Accepted solution]

The only valid option I can think of is to use the rex command with mode=sed to eliminate this part from the email address:

... | rex field=mail mode=sed "s/\+\w+@/@/g"

[Simon]

Thanks, seems like there's no other possibility.
I also got an answer from support that splunk doesn't replace more than one matching group in transforms.conf

[ftk]

You should be able to do this:

rex "\<(?<realfrom>\S+)\+\w+@"

I tested this with this search on my system:

* | head 1 | eval blah="from=<[email protected]>" | rex field=blah "\<(?<realfrom>\S+)\+\w+@" | table blah realfrom

[ftk]

Oh ok. In that case ziegfried is right, you'll want to use rex in sed mode.

[Simon]

Hey, thanks for your answer, but it's important that I get the domain name (newsletter.domain.com) in my match too. Only using the username of the email address for identifying senders is not distinct enough.