Exposure Validation Finder

Lesson 4 of 8

Scoping your first validation

A first validation goes well when the scope is small, the targets are named, and everyone who might see an alert knows it is coming.

Start with a question, not a product

Write down the one question the first validation must answer. Good first questions are narrow. "Does our EDR stop the credential theft techniques used against our sector?" "Can a phished user laptop in the sales VLAN reach any server holding customer data?" "Does the new email gateway filter the attachment types the old one caught?"

A narrow question tells you which tool type to pilot, which hosts are involved, and what success looks like. "Validate our security posture" tells you none of those things and produces a report nobody acts on.

Name the targets

For path questions, name two or three target assets: a domain controller, a payment database, a source code repository, a cloud account with production data. A path test without targets will find something, but not necessarily the thing you care about.

For control questions, name the controls by product and policy: the EDR policy on finance laptops, the outbound web proxy for the head office, the SIEM rules for lateral movement. A control test that runs against "the endpoint" can pass on a well-protected test VM and say nothing about the fleet.

Pick starting points that look like real attacks

Most validation assumes a foothold. Choose footholds that resemble how attackers actually get in. An ordinary user workstation is the most common. A server in the DMZ and a contractor's VPN connection are others.

Avoid starting from an administrator's machine for the first test. It will find a lot, quickly, and the findings will mostly say that administrators have administrator rights.

For agent-based tools, the starting point is where you install the agent, test point or simulator. For agentless automated pentesting, it is where the scanning host sits on the network, plus any credentials you choose to give it. Decide in advance whether the tool starts with no credentials, a standard user account, or both.

Write rules of engagement

Even tools sold as production-safe need written rules. One page is enough. It should cover:

Ask the vendor to fill in the "what the tool may do" section from its own documentation. The answer shows how well the vendor understands its own safety controls.

Tell the right people

A validation run will trigger alerts if it is working. Decide in advance whether the SOC knows. There are good reasons both ways. Telling them avoids wasted investigation. Not telling them tests detection and response as they really are.

A common compromise is to tell the SOC lead and not the analysts, with a code word to stop escalation. Whatever you choose, tell the owners of the target systems, the network team and anyone who approves changes. Surprises in production end pilots early.

Decide what you will measure

Before the first run, write down what a result looks like. For control tests: the share of techniques prevented, detected and missed, per control. For path tests: whether each target was reached, in how many steps, and which steps were detected. Lesson 7 covers which of these numbers hold up over time.

Agree also what will happen after. A finding with no owner is a finding nobody fixes. Name who receives control failures, who receives path findings and when they will meet to go through them.

Keep the first pilot short

Two to four weeks is enough for a first pilot. Week one installs and runs a baseline. Week two fixes one or two findings and retests. The remaining time shows how the tool behaves on a schedule and how the vendor responds to questions.

The results worksheets on this site each list what to check in a pilot for one tool type. Print the one that matches your question and take it into the vendor meeting.

A small example

A retailer's first question is whether a compromised store-office PC could reach the card payment network. The targets are two payment servers. The starting point is one store-office PC with a standard user account. The rules forbid exploits against the payment servers themselves and allow everything else in the store VLAN during closing hours. The SOC lead knows; the analysts do not. Success means a clear yes or no for each server, with the path if yes.

That scope points to automated pentesting or executed attack path testing, run agentless from one host. It fits in one page and one week. Most first validations should look about that size.


Previous: Where validation fits in CTEM. Next: Agents, agentless and production safety.