When the Attack Is No Longer Something You Read About

There is a big difference between reading about a cyberattack and being involved when one is actually happening. When you read a security report, everything is neatly presented: the attacker gained access, the malware executed, the indicators were identified, and the recommended controls are listed at the end. When you are in the middle of an incident, there is nothing neat about it. You have incomplete information, systems that you normally trust are suddenly questionable, and every decision has consequences.

I recently found myself on the other side of that line, involved in responding to a real security incident involving a remote-management environment. I am deliberately leaving out the organisation, systems, addresses, usernames and other identifying information. This is not about the client. It is about the experience, the lessons and the things that become very real when you are no longer looking at an incident from the outside.

The first thing that changes is trust

Remote-management platforms exist because we trust them. They are designed to connect administrators to servers and workstations, execute commands, transfer files and manage large numbers of devices from one place. That is exactly what makes them so useful to an IT team, and exactly what makes them so attractive to an attacker.

During the response, the question was no longer simply whether the remote-management platform was functioning correctly. The much more uncomfortable question became: can we still trust it?

That changes your thinking completely. A command executed through a management platform can look exactly like something an administrator would normally do. The same system that makes your job easier can become the mechanism through which an attacker reaches everything else.

You stop thinking about availability first

In normal IT operations, our instinct is to restore services as quickly as possible. During an incident, that instinct has to be controlled.

Stopping a management platform that administrators depend on is disruptive. Keeping a potentially compromised management platform running can be far more dangerous.

Containment therefore becomes more important than convenience. The platform was stopped while the environment was investigated, and components were considered separately rather than simply bringing everything back online at once. The principle was simple: restore only what you need to investigate, not everything you need to operate.

That is not always an easy decision when people are waiting for systems to come back.

The logs suddenly mean something different

Before an incident, a session registration, connection event or command execution may simply be another entry in a database.

During an investigation, those same records become pieces of a story.

Timestamps, source addresses, machine names, logged-on users, client versions, connection history, session activity and technician actions all start becoming important. One unusual event may mean nothing. Several unusual events occurring together can completely change the picture.

This is where good logging stops being a compliance exercise and becomes part of your ability to understand what actually happened.

And there is another uncomfortable lesson: “not confirmed infected” does not necessarily mean “clean.”

Some systems may have confirmed malicious execution. Others may show attempted activity, interaction or commands where successful execution cannot be established. Those distinctions matter, because incident response is not about pretending we have certainty when we do not.

Then you start looking at the network differently

One of the biggest lessons for me was how quickly a firewall changes from being a traffic-control device into an investigation tool.

When you discover that a privileged server can communicate freely with the Internet, you start asking a very different question.

Not:

“Does this server need Internet access?”

But:

“What Internet access does this server actually need?”

That led to a much more restrictive model: allow the specific destinations and services that can be justified, and deny the rest.

It sounds obvious when written down. In a live environment, it requires careful investigation because legitimate applications have dependencies, licensing services need connectivity, authentication may require external services, and remote-management platforms have their own communication requirements.

The important lesson is not to blindly block everything. It is to understand what should be communicating, why it should be communicating and what process is responsible for that communication.

Defence in depth stops being a buzzword

We often talk about defence in depth in architecture documents.

During an incident, you experience it differently.

Endpoint protection, firewall controls, DNS filtering, malicious IP blocking, FQDN-based controls, restricted server egress, centralised logging and monitoring all become pieces of the same defensive picture.

No single control is guaranteed to stop an attacker.

That is not failure.

The objective is that if one control is bypassed, another control creates friction, and somewhere along the way the activity generates evidence that allows you to see what is happening.

What I took away from being inside the response

The biggest lesson was not about one product, one vulnerability or one particular attacker.

It was about trust.

As IT professionals, we build systems around trusted relationships. We trust our management platforms. We trust our administrators. We trust our firewalls. We trust our authentication systems. We trust that a legitimate management command is exactly what it appears to be.

A serious incident forces you to question those assumptions.

It also reminds you that recovery is not simply switching the server back on.

Recovered does not automatically mean trusted.

Before a privileged platform returns to normal operation, you need to know that suspicious registrations have been dealt with, credentials and sessions have been reviewed, unnecessary access has been removed, firewall rules have been tightened, logging is available outside the affected system and the organisation has a reasonable chance of detecting the same behaviour if it happens again.

The view from Expert in the Cloud

This experience changed the way I look at some of the architecture decisions we make every day.

We spend a lot of time designing resilient networks, secure firewalls, cloud environments and management platforms. But security is not only about how well those systems work when everything is normal.

It is about what happens when something goes wrong.

What can this system reach?

What can reach this system?

What happens if this trusted application is compromised?

And, perhaps most importantly, would we see it?

Those questions are easy to ask when you are sitting comfortably reviewing an architecture diagram.

They feel very different when you are asking them in the middle of an actual incident.

That is probably the biggest lesson I took away.

Cybersecurity is not just about protecting systems from attackers. It is about designing environments that remain understandable, controllable and observable when the assumptions we normally rely on suddenly disappear.

That is what being in the middle of an attack teaches you that reading about one never can.

— Expert in the Cloud

Be the first to comment

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.