Welcome to electus’s documentation!

Getting started

Electus is a package for generating alerts based on combinations of indicators. It allows the user to create a library of flexible behavioural signatures that can be reused in different contexts. Events can come from anywhere as long as they are python dictionaries. For example, this could be events fetched from a REST API or a host based tool such as phishermon. To install electus run:

pip install electus

Once the signatures have been defined they can be combined in different ways (called jobs). For example, one type of job is a Sequence which looks for events in a specific order within a given time window.

A very basic example is

from electus import Electus

# Create some sample data
events = [{
           "process": "winword.exe",
           "parent_process": "explorer.exe",
           "datetime": "2017-10-16T17:15:47.030114"
          {"process": "powershell.exe",
           "parent_process": "winword.exe",
           "datetime": "2017-10-16T17:15:48.030123"

# The library.json file stores the indicator definitions
# The combinations of indicators are defined in jobs.json
e = Electus(library_conf='library.json', job_conf='jobs.json')

for event in events:
    alerts = e.evaluate_event(event)
    if alerts:

Defining a library of indicators

Indicators are the fundamental unit within electus. Given an event with certain fields, they are used to match specified key-value pairs.

Indicators are defined in a json file (typically called library.json). For example a very basic library.json (with one indicator) looks like

    "datetime_field": "datetime",
    "indicators": [{"field": "process", "value": "powershell.exe"},
                   {"field": "parent_process", "value": "winword.exe"}]

This will match events where the process field is “powershell.exe” and the “parent_process” field is “winword.exe”. Each entry in the library requires a unique name (“word_launching_powershell” in this case) and the following fields:


This is a list of dictionaries with the following fields:

field:The key to look for in the event object.
value:The value to match on in the event object.
match:(Optional) If false the indicator will only hit if the data does not match the specified value. Can be used to eliminate false positives. Defaults to true.
regex:(Optional) Specifies if the value parameter is a regex (otherwise a straight comparison will be used). Defaults to false.
case_sensitive:(Optional) Whether or not to do case sensitive comparisons. Defaults to true.

(Optional) A score to assign if all of the indicators match. Used in the Filter job.


(Optional) This is the key in the event used to compare timestamps where necessary. Required for Sequence jobs.


The name can only contain alphanumeric characters (plus underscores) and cannot start with a number. This is due to the way the boolean logic is parsed for join jobs.

A larger example library.json is shown below:

  "word_launching_powershell": {"datetime_field": "datetime",
                                "weight": 10.0,
                                "indicators": [{"field": "process", "value": "powershell.exe"},
                                               {"field": "parent_process", "value": "winword.exe"}]

  "temp.exe": {"datetime_field": "datetime",
                                "indicators": [{"field": "process", "value": ".*temp.exe", "is_regex": true, "case_sensitive":false}]
  "bad.exe": {"datetime_field": "datetime",
                                "indicators": [{"field": "process", "value": "bad.exe"}]
  "abnormal_svchost": {"datetime_field": "datetime",
                            "indicators": [{"field": "process", "value": "svchost.exe"},
                                           {"field": "parent_process", "value": "services.exe", "match": false}]

In the above example, the “abnormal_svchost” indicator will only hit if the parent process of svchost.exe is NOT services.exe.

Combining indicators

Jobs are used to combine sets of indicators to alert on events of interest. The job definitions are typically stored in jobs.json. The different job types are:


This can be used to combine indicators with boolean logic expressions. For example, the expression “A and (B or C)” will alert if A and B or A and C are seen.

Below is a basic example of using a join job

"join_job_example": {"type": "join",
                     "features":["temp.exe", "bad.exe"],
                     "join_expression": "temp.exe and bad.exe"}

The following fields are required:

type:This is the type of job (in this case join)
features:This is a list of indicator names specified in library.json
 This is the logical statement for how to combine the indicators. AND, OR, &&, || and parentheses are supported.


Each definition in the library file can be weighted via the “weight” key. The filter job sums up all of the weights and alerts if it exceeds a specified threshold.

The filter job looks similar to the join job:

"winword_launching_powershell": {"type": "filter",
                                 "threshold": 10,
                                 "features": ["word_launching_powershell"]}
type:The type of job (in this case filter)
threshold:This is the minimum score for alerting. In this case the sum of scores must be >=10
features:This is a list of indicator names specified in library.json.


This alerts if an orded sequence of indicators occurs within a time window. For example, only alert if you see A then B then C within 10 seconds.

In this example we are looking for temp.exe then bad.exe within a span of 5 seconds.

"temp.exe then bad.exe": {"type": "sequence",
                          "features":["temp.exe", "bad.exe"],
                          "sequence": ["temp.exe", "bad.exe"], "time_window": 5}
type:The type of job (in this case sequence)
sequence:This is the list of indicator names from the library in the order they should be seen
features:This is a list of indicator names specified in library.json.
time_window:This is the total time window (in seconds) to look for the sequence in.