You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
I have some issues around the "date" and "time" fields which come from syslog, either as pipe-delimited fields or as mapped JSON fields.
The fundamental question is: are these fields actually used for anything, apart from being included in alert messages? And does it matter if the format is not YYYY-MM-DD or HH:MM:SS?
That is: both fields are being set to the same value! According to syslog-ng documentation this field contains the BSD format date/time (e.g. Jun 13 15:58:00).
It looks like it should be possible to get syslog-ng to synthesise the required date and time using value-pairs. But as I'm using rsyslog rather than syslog-ng I don't have much interest in developing this.
What I am interested in is the importance of these fields, which affects how I configure rsyslog.
rsyslog's default JSON format generates a single ISO format string (e.g. 2019-10-28T16:56:43.171226+00:00). Should I put that in DATE, and leave TIME undefined? Or something else? Does it matter?
Note: with some effort, it's also possible to get rsyslog to generate custom JSON with YYYY-MM-DD and HH:MM:SS - I have documented this here in PR #146. The question is, is this necessary and/or worthwhile?
(2) The pipe examples shows the "date" field in YYYY-MM-DD, but this appears to be inconsistent with alerts which report MM-DD-YYYY. e.g.
[**] [1:5000077:3] [OPENSSH] Attempt to login using a denied user [**]
[Classification: unsuccessful-user] [Priority: 1] [127.0.0.1]
[Alert Time: 10-28-2019 16:18:19.654404]
2019-10-28 16:18:19 192.168.2.1:514 -> 192.168.2.1:22 user notice sshd
Message: User ubuntu not allowed because shell /etc/passwd is not executable
[Xref => http://wiki.quadrantsec.com/bin/view/Main/5000077]
(compare line 3 and line 4 above). This is not locale-dependent; the alert time format comes from CreateTimeString in src/util-time.c
It appears to be legacy from poor design choices made in Suricata / Snort
Describe the solution you'd like
Document what function the syslog "date" and "time" fields perform
Describe alternatives you've considered
Change date to MM-DD-YYYY for consistency? No, YYYY-MM-DD is better for many reasons
Additional context
N/A
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
I have some issues around the "date" and "time" fields which come from syslog, either as pipe-delimited fields or as mapped JSON fields.
The fundamental question is: are these fields actually used for anything, apart from being included in alert messages? And does it matter if the format is not YYYY-MM-DD or HH:MM:SS?
Context
(1) I notice from
sagan-rules/json-input.map
That is: both fields are being set to the same value! According to syslog-ng documentation this field contains the BSD format date/time (e.g.
Jun 13 15:58:00
).It looks like it should be possible to get syslog-ng to synthesise the required date and time using value-pairs. But as I'm using rsyslog rather than syslog-ng I don't have much interest in developing this.
What I am interested in is the importance of these fields, which affects how I configure rsyslog.
rsyslog's default JSON format generates a single ISO format string (e.g.
2019-10-28T16:56:43.171226+00:00
). Should I put that in DATE, and leave TIME undefined? Or something else? Does it matter?Note: with some effort, it's also possible to get rsyslog to generate custom JSON with YYYY-MM-DD and HH:MM:SS - I have documented this here in PR #146. The question is, is this necessary and/or worthwhile?
(2) The pipe examples shows the "date" field in YYYY-MM-DD, but this appears to be inconsistent with alerts which report MM-DD-YYYY. e.g.
(compare line 3 and line 4 above). This is not locale-dependent; the alert time format comes from
CreateTimeString
insrc/util-time.c
It appears to be legacy from poor design choices made in Suricata / Snort
Describe the solution you'd like
Describe alternatives you've considered
Change date to MM-DD-YYYY for consistency?No, YYYY-MM-DD is better for many reasonsAdditional context
N/A
The text was updated successfully, but these errors were encountered: