An advanced persistent threat, or APT, is an adversary with sophisticated expertise and significant resources who pursues its goals repeatedly over a long period and adapts when defenders push back. NIST's definitions focus on capability and persistence. None of them depends on knowing which country or group is behind the activity.
Three words, three tests
NIST publishes the definition in several documents, and the wording shifts a little between them. Read together, they give three tests.
Advanced. SP 800-30 Rev. 1 describes an adversary with "sophisticated levels of expertise and significant resources". It also says such an adversary uses several attack channels, for example cyber, physical and deception, to create chances to reach its goals.
Persistent. SP 800-171 Rev. 3 says the adversary "pursues its objectives repeatedly over an extended period of time". The same text says it adapts to defenders' efforts to resist it.
Threat. The adversary has intent and the means to act on it. SP 800-39 and SP 800-53 Rev. 5 describe the aims as establishing and extending footholds in IT infrastructure, either to steal information or to undermine critical operations.
What the definitions leave out
Notice what is missing. No definition says "a nation state". No definition names a malware family, an exploit type or an industry. A more recent NIST publication, SP 800-172 Rev. 3, widens the idea further. It describes an adversary or a collection of adversaries who may be collaborating, opportunistically overlapping, or inadvertently converging.
That matters for how you talk about threats inside your company. If your risk register says "APT", you are describing a level of effort and patience, not a named villain. The practical question is how much effort a determined attacker would spend on you, and what would stop them at each step.
Why the label gets misused
Vendors and news stories use "APT" as shorthand for "scary state-backed group". That use is common but it is not the NIST meaning, and it causes two mistakes.
The first mistake is dismissal. A team decides "we are not interesting to a government" and skips the whole topic. The definitions do not require a government. A well-funded criminal crew that returns for months can meet the same tests.
The second mistake is fatalism. A team decides nothing can stop a sophisticated adversary and stops investing in basics. The advisories listed below show the opposite: many of the recommended defences are ordinary, such as patching internet-facing systems, phishing-resistant MFA, least privilege and central logging.
How to use the definition
- Write your threat scenarios in terms of behaviour: long dwell time, credential theft, abuse of trusted tools, return after cleanup.
- Ask whether your detection would still work if the attacker used no malware at all.
- Plan for the second visit. Persistence means removal of one foothold is not the end of the incident.
- Keep attribution out of the first hour of an incident. You can act on behaviour without knowing who it is.
You can apply these without buying anything. Safeguard's reachability analysis and dependency scanning cover one slice of the problem, the software you ship, and you can try them with the free scan at https://gold.safeguard.sh/scan. Identity, logging and network controls need their own owners.
Quick check
Does an APT have to be a government?
No. The NIST definitions describe expertise, resources and persistence. They do not require a state sponsor.
Is one long intrusion enough to call it an APT?
The definitions describe repeated pursuit over an extended period, so a single short event does not fit well. Treat the label as a planning tool, not a verdict.
Sources
Last reviewed: 19 October 2026.
