Security Operations & Resilience

Endpoint Detection and Response (EDR) Basics

what an endpoint agent sees, how it decides, and what you can do with it

2 min readupdated by Syed Khaleel Ahmed

Traditional antivirus mostly asks "does this file match something known to be bad?" Endpoint detection and response (EDR) asks a bigger question: "what is this system doing, and does it look like an attack?" It is now a core tool in most security operations.

What an EDR agent records

A lightweight agent on laptops and servers continuously records activity:

  • Process creation, with the parent process and full command line
  • File creation, change and deletion
  • Registry changes (on Windows)
  • Network connections
  • Logins, driver loads and use of scripting tools

This telemetry is stored centrally, so analysts can search it later, even for activity that was not flagged at the time.

How it detects

  • Signatures and known-bad indicators for familiar malware.
  • Behavioral detection: patterns such as a document launching a shell that downloads a file, or a process dumping credentials from memory. These are often mapped to MITRE ATT&CK techniques.[1]
  • Machine learning to score unfamiliar files or behavior.
  • Threat intelligence matches such as known malicious domains.

What responders can do

  • Isolate a host from the network while keeping a connection to the console.
  • Kill a process or quarantine a file.
  • Collect files and a forensic triage package, or open a remote shell to investigate.
  • Search the whole fleet for the same indicator to see who else is affected.
  • Roll back some changes, depending on the product.

Limits to keep in mind

  • Coverage gaps: unmanaged devices, old systems and some operating systems may have no agent.
  • Tampering and evasion: attackers try to disable or blind agents, so tamper protection and alerts on agent health matter.
  • Alert noise if not tuned.
  • It sees endpoints, not everything: it will not see network devices or cloud services on its own.
  • It needs people. Tools detect; analysts decide.

EDR, XDR and MDR

  • XDR (extended detection and response) pulls in more sources, such as email, identity, network and cloud, and correlates them with endpoint data.
  • MDR (managed detection and response) is a service in which an outside team monitors and responds for you.

Independent exercises such as MITRE's ATT&CK Evaluations show how products detect emulated attacks, but they are one input among many.[2]

Questions to ask of an alert

Which user and process? What launched it? What did it touch and where did it connect? Is this normal for this machine? Who else shows the same pattern?

Sources

  1. MITRE ATT&CK
  2. MITRE Engenuity ATT&CK Evaluations

Test yourself

My notes

Private — saved only on this device, never sent anywhere.