SOC Ops Rocks

I find that the vast majority of vendor integrations and playbooks automations are 90% identical: ingest inbound data (array, object, etc), parse through it whilst validating and extracting, then finally pushing it out.

This means I use the same code aaaaaall the time. So I thought I would make a little cheatsheet for the basics (to prevent me googling the same things over and over). This list will change/grow over time.

Not covered in this cheatsheet:

• Local file handling (open, read, close, etc)
• HTTP and SSL request/replies
• Time/date handling

To keep formatting simple, “<<tab>>” respresents a real Tab in the code

#	***	***
#	Common Lib Imports
	import json, re, pprint, time, random, base64
#	Basic string
	myString = “hi”
	myString += ” I added a bit”
#	Convert
	myNum = int(myString)
	myInt = str(myNumber)
#	If
	if varA > 5: <<tab>>print(“High”) elif varA > 0: <<tab>>print(“Medium”) else: <<tab>>print(“Zero”)	Basic structure Remember indents are tabs
	if myString == “compare me”:	Simple string comparison
	if myVar is None:	None / Null
	if not myVar:	Check if value exists, but is empty
	if (a == b) and (not b == c):
	if (a < 5 < b):	If both are true
	if isinstance(myVar, list):	list, dict, str, int… etc
#	Lists (arrays)
	myList = []
	myList.append(“bob”)
	[output1, output2] = myList.split(‘character’)
	myList = myString.split(” “)
	myString = “-“.join(myList)
#	Python Dict
	myDict = {}	alternative >> myDict = dict()
	myVar = myDict.get(“key”, default_value)	Extract a value into another var
	myVar = myObject[‘key’]	Similar to above, but will return error if not found
	myDict[‘key’] = ‘value’	Set a new value to the Dict
	if myKey in myDict:	Check if key exists
	myList = myDict.keys()	Same for “.values()”
	my_dict.pop('key', None)	Safe is key exists or not
	del my_dict['key']	errors if key doesn’t exist
#	Json	(Technically not a Dict, but handled very similarly)
	import json	Import to code to use the calls
	myJson = {}
	myJson = json.loads(string)	Where string is in the form  “{‘key’:value}”
	myStr = json.dumps(string)
	myString = myJson[‘key’]
	myJson[‘key’] = newValue
	myJsonAsString = json.dumps(myJson)
	if myKey in myJson:
#	Loops
	for value in myList: <<tab>>print(value)	Loop a List
	while condition == True:	Simple while
	for myInt in range(5):	myInt will be 1,2,3,4,5
	for index in range(len(myList)):	index will represent a number of the position in array
	for key, value in myJson.items():	Loop through Json/Dict
	for key in jsonObject: <<tab>>value = jsonObject[key]	Alternative to the above
	break	Break out of the current loop (1 layer) to next code
	continue	Stop processing this loop, and go to next iteration of this loop structure
#	Regex
	newString = re.sub(r”pattern”, “replaceWith”, targetString))	regex substitution
	arrayResults = re.findall(r’pattern’, targetString)	regex findall
	matches = re.match(r”Goodbye”, “Hello”) if match is None:	regex match and test
	objectResults = re.search(“[a-z]+”,myVar)	Returns complex output
#	base64
	encoded = base64.b64encode(‘Hello World’)	Encode a string into Base64
	readableData = base64.b64decode(encoded)	Convert Base64 back into the original string
#	Try
	try: <<tab>>print(nonExistentVariable) except: <<tab>>print(“Something went wrong”) finally: <<tab>>print(“The ‘try except’ is finished”)	Remember tab indents
#	Debug
	print()	Simple print
	pprint()	Print more complicated objects from pprint import pprint
#	Bits and Bobs
	random.seed() myInt = random.random()	Int between 0.0 and 1.0
	randomInt = random.randint(1,100)	Int between x and y
	time.sleep(1)
	print(type(x))	To get the Type of a variable

Last week I had an intersting chat with a security team:

• Our workload is very unpredictable
• We want SOAR to intelligently auto-prioritise incidents
• And when we are ‘busy’ auto close low priority tickets
• but we still want automated IOC enrichment, full auditing, etc

Coupled with intelligent prioritising this is a great idea

Request : “if workload is high, auto close incident”

• After a new incident workflow enriches, we calculate the current team workload
• For every open incident: Priority1 = 4 points, open Priority2 = 3 points, etc
• If total points is >20 then auto close the incident with a note “auto closed due to too much workload”

This is great, but I see an improvement. Workloads change very quick, you might be busy right now, but in 1 hour everything gets resolved, then you have no tickets to look at.

My alternative: “create, enrich, wait, auto close”

• Any low priority incident starts a 3 day timer
• Incidents are assigned to the team, not an individual
• If an analyst has capacity then can self assign and now own the ticket
• If the incident isn’t touched in 3 days it is auto closed
• We create dashboards that look at the incident count per close duration
• This dashboards show how many incidents / type are closed without being looked at

I’m an ex-analyst, I know that low quality alerts can contain valuable information, we don’t always have the time, but that ticket still needs enrichment for future analysis if we need to come back to it.

At least using SOAR for automation you ensure that:

• The incident was logged
• The details were enriched
• You were able to reach out to members of the company to validate
• Auto log all information/decisions for future audit and reviews
• The playbook had the option to double check the alert is low priority (and self re-prioritize if not)

…which is significantly more than I was able to control a few years ago 🙁

Andy

A man walks into a bar, ouch

This is a quick ‘joke’, it takes 2 seconds to say, everytime I tell it. I have no concerns giving a SLA for this joke. On the other hand…

Knock Knock…

Whilst I know *I* can tell this joke in under 5 seconds I’m entirely relying on the person I’m talking to, is it representative to apply a SLA on me?

Compare this to a SOAR playbook: any local task we have control over, but it’s not so simple when we wrap a business process around this:

• Any interaction that involves human input (especially where that person is not part of our team, and we can’t kick them)
• A query that potentially takes hours to complete
• Unstable technology we can’t change
• Technology belonging to another team

So how do we apply such SLA to playbooks ?

SLA for an entire Incident

Pro – Quick to configure. Great for small simple playbooks.

Con – Very inflexible.

A timer starts with the incident, if the ticket takes longer, we have a SLA breach.

• 

SLA for each individual task

Pro – Finely tuned

Con – Administrative overheads building and maintaining

Start a timer for each specific task, if that task takes too long we can either alert, skip the task, or take a different playbook route and escalate the process to the senior team.

Timed Section

Pro – Flexible. quick to deploy

Con – none?

E.g. Task 1 starts timer, task 5 pauses it, task 7 resumes it, task 10 closes it.

• 

• 

Knock Knock (including SLA)

• The “Joke SLA” represents the entire incident
  • Terminology “Incident SLA”
• The “My Team SLA” stops and starts
  • Terminology “Timer”
• The “Punchline SLA”
  • “Task specific SLA”

Andy

That moment when you realise SOAR is just a bunch of building blocks.

Andy

After recently getting married, I quickly discovered my wife was cheating on me, with these three:

As soon as I left the house to go shopping, seeing the family, or working away my (NOT-SO-)good lady would jump into bed, turn on the TV and spend quality time watching series without me, getting ahead in a series we were watching together !

This is disastrous and would no doubt lead to a playbook on how get divorced. Urgent actions were needed, so I turned to SOAR!!

(Really this is a blog post about automating the whitelisting/blacklisting of IP/domains. Either for a SOC team who are detecting new attacks, or whether it’s members of staff managing their own policies. But hey I love drama, pun intended. Read to the end where I discuss “This In Business”)

I need a process that:

• Automatically detects this activity
• Automatically remediates this activity
• Validate if I’m home
• Communicates with the culprit (played by my hussy wife)
• Seeks confirmation from sysadmin (in this post the victim is played by the flawless handsome and honorable me)
• Audit logs, SLA, etc

Setup

SOAR running on a virtual machine at home

The DNS activity/alert is generated by PassiveDNS on a Raspberry PI on a network sniffer port

My home network has lots of Unifi in. Unifi do amazing kit with a full API available to control the WAP, Firewall rules, network config etc. I <3 the Unifi!

Workflow Process

To avoid a SIEM at home, I simply have PassiveDNS forward logs for Netflix DNS requests direct to SOAR.

PreProcessing is then used to make sure that all prior tickets/incidents are closed (i.e. check this is a new situation)

SOAR then queries my Unifi controller to see if my personal mobile phone is connected to the WIFI. If “yes” then I am home, if “no” I’m likely out travelling.

The VICTIM (innocent me) who is likely in a hotel/shop is then either sent an email with a choice to block the activity instantly, or to request a justification….

….or if I’m being an uber admin, I can use the mobile app to decide….

If I chose to ask “EnforceJustification” a questionnaire is send to the wicked one!

Answers are forwarded to the sysadmin

Of course being the benevolent kind generous soul I am, I of course decide to Allow this traffic (between you and me, I watched this the other day, so I’m already ahead of her… #guilty)

And thus our marriage is saved. Should I train to become a marriage counsellor?

SOAR Value Realised

I previously talked about the value of SOAR and I think this playbook ticks off many of those:

• Reduce alert overhead
• Quicker to act
• Standardise your workflow
• Standardise approvals
• Revitalise legacy/simple tools

This In Business

Many of us are familiar with this phone call from a member of staff who was denied a work related website:

“How dare you block my access to the internet, I need that website to do my job! You’re stopping business! Stop being paranoid! I’m going to report this!”

So let’s adapt the above process:

• A workflow initiated by the end user (through a ticketing system, SIEM, emails, or other)
• Playbook asks the user which domain to whitelist, for how long, and for what justification
• Playbook then checks if this domain was requested before. If denied, ticket can be closed. If approved a few times should we consider a long term whitelist?
• Playbook then enriches the domain against Threat Intel: Is it known malicious, is it less than a week old, is it an inappropriate category of site
• Playbook enriches with ActiveDirectory to determine that users manager
• Playbook emails manager for approval with all the details and buttons of “yes approve” or “no, block”
• If the manager approves the domain is whitelisted
• And after the correct amount of time, automatically removes the policy change to clean up.

We now have a process that is operational 24/7, works at the speed of the affected staff (not the huge workload of the security team), takes no effort on your team, does threat enrichment and sanity checking, cleans up after itself, has SLA, RBAC, is fully audited. All the while, no member of staff was given access to any security tool!

Useful ?

Andy