SPO Automation

This package contains a script to automate parts of workflows for SPO.

Requirements

Python 3.10 or newer

Setup

Install the package using pip:

pip3 install git+ssh://git@gitlab.local.northwave.nl/mneikes/spo-automation.git

Alternatively, for local development, clone the repo and install it as editable package:

git clone git@gitlab.local.northwave.nl:mneikes/spo-automation.git
pip3 install -e .

Then, create a config file config.ini with the following content:

[jira]
host = https://jira.local.northwave.nl
username = spo-workflow-automation
password = (1Password)
dryrun = yes

[openai]
key = (OpenAI key)

The password for the spo-workflow-automation user is currently shared with Jair and Moritz. The OpenAI key can be found by clicking the "View Code" button in the Azure AI studio chat playground.

The dryrun option controls whether the code should run in passive mode, and only log what it would do (dryrun = yes), or whether it should actually make changes to the Jira tickets that it processes (dryrun = no).

Usage

Once installed (see Setup above), run the script like so:

usage: nw-spo-automation [-h] {issue,search-issues} ...

Apply automation to SPO Jira tickets

positional arguments:
  {issue,search-issues}
                        Specify the mode in which to operate
    issue               Run automation for a specific issue
    search-issues       Search for applicable issues and run automation on all of them

options:
  -h, --help            show this help message and exit

Individual tickets

To run automation for a specific ticket, use the issue subcommand:

usage: nw-spo-automation issue [-h] --key KEY

options:
  -h, --help  show this help message and exit
  --key KEY   The Jira issue key

Search for issues

To search for all applicable issues and run automation on all of them, use the search-issues subcommand like so:

usage: nw-spo-automation search-issues [-h] [--limit LIMIT]

options:
  -h, --help     show this help message and exit
  --limit LIMIT  Limit the number of issues that will be handled

How it works

The SPO automation currently operates in two phases:

Classification
Enrichment

During classification, the email content of a ticket is processed to map it to one of a predefined set of subcategories. These subcategories are grouped into security- and privacy-related topics.

The list of categories can be found in ./spo_automation/ticket_classifier.py.

If there is an enrichment function defined for the subcategory, that function will then be run with the ticket.

The enrichment functions can be found in ./spo_automation/workflows. They are grouped into security and privacy functions.

Adding a new workflow

If necessary, expand the classifications in ./spo_automation/ticket_classifier.py

Then, create a new file spo_automation/workflows/<category>/<subcategory>.py, using the category and mapped subcategory from the ticket classifier.

The mapped subcategory is the machine-friendly version of a subcategory name. For example, the subcategory "Data Breaches (e.g., incorrectly sent email)" maps to the mapped subcategory data_breaches. This mapping happens explicitly in ./spo_automation/ticket_classifier.py.

Your new workflow file must have at least a function with the following signature:

def on_create(spo_automation, issue):
    pass

This function will be called with all issues that fall under your new subcategory. Example:

from spo_automation import openai_wrapper, templating

system_message = """
You are an AI assistant As an AI assistant to a Security and Privacy Officer.
Your task is to [...]

Respond only in JSON format with fields for ...

Example:
{
  "key": "value",
  # give a specific example here
}
"""

email_template = """
The AI concluded the following: {{ resp.key }}
It found the following issues:
{% for issue in resp.issues %}
 - {{ issue }}
{% endfor %}
"""

def on_create(spo_automation, issue):
    openai_instance = openai_wrapper.OpenAIWrapper(
        spo_automation.config.openai_key,
        system_message,
    )

    email_content = issue.get_field("SPO: E-mail Body")
    resp = openai_instance.create_json_completion(email_content)

    response_email = templating.render_template(
        template = email_template,
        context = {
            "resp": resp,
        }
    )

Skixie / Phishing-GPT