Appendix P: Time functions
editAppendix P: Time functions
editThe time functions detect events that happen at unusual times, either of the day or of the week. These functions can be used to find unusual patterns of behavior, typically associated with suspicious user activity.
The machine learning features include the following time functions:
- You cannot create forecasts for anomaly detection jobs that contain time functions.
-
The
time_of_dayfunction is not aware of the difference between days, for instance work days and weekends. When modeling different days, use thetime_of_weekfunction. In general, thetime_of_weekfunction is more suited to modeling the behavior of people rather than machines, as people vary their behavior according to the day of the week. -
Shorter bucket spans (for example, 10 minutes) are recommended when performing
a
time_of_dayortime_of_weekanalysis. The time of the events being modeled are not affected by the bucket span, but a shorter bucket span enables quicker alerting on unusual events. - Unusual events are flagged based on the previous pattern of the data, not on what we might think of as unusual based on human experience. So, if events typically occur between 3 a.m. and 5 a.m., an event occurring at 3 p.m. is flagged as unusual.
- When Daylight Saving Time starts or stops, regular events can be flagged as anomalous. This situation occurs because the actual time of the event (as measured against a UTC baseline) has changed. This situation is treated as a step change in behavior and the new times will be learned quickly.
Time_of_day
editThe time_of_day function detects when events occur that are outside normal
usage patterns. For example, it detects unusual activity in the middle of the
night.
The function expects daily behavior to be similar. If you expect the behavior of
your data to differ on Saturdays compared to Wednesdays, the time_of_week
function is more appropriate.
This function supports the following properties:
-
by_field_name(optional) -
over_field_name(optional) -
partition_field_name(optional)
For more information about those properties, see the create anomaly detection jobs API.
Example 1: Analyzing events with the time_of_day function.
{
"function" : "time_of_day",
"by_field_name" : "process"
}
If you use this time_of_day function in a detector in your anomaly detection job, it
models when events occur throughout a day for each process. It detects when an
event occurs for a process that is at an unusual time in the day compared to
its past behavior.
Time_of_week
editThe time_of_week function detects when events occur that are outside normal
usage patterns. For example, it detects login events on the weekend.
The time_of_week function models time in epoch seconds modulo the
duration of a week in seconds. It means that the typical and actual values
are seconds after a whole number of weeks since 1/1/1970 in UTC which is a
Thursday. For example, a value of 475 is 475 seconds after midnight on
Thursday in UTC.
This function supports the following properties:
-
by_field_name(optional) -
over_field_name(optional) -
partition_field_name(optional)
For more information about those properties, see the create anomaly detection jobs API.
Example 2: Analyzing events with the time_of_week function.
{
"function" : "time_of_week",
"by_field_name" : "eventcode",
"over_field_name" : "workstation"
}
If you use this time_of_week function in a detector in your anomaly detection job, it
models when events occur throughout the week for each eventcode. It detects
when a workstation event occurs at an unusual time during the week for that
eventcode compared to other workstations. It detects events for a
particular workstation that are outside the normal usage pattern.