article · 2026-08-12 · Elena Cho

From IOC to behavior: retiring indicator-only detection

Most intel programs still measure themselves in indicators shipped. The programs that reduce incidents measure coverage that survives infrastructure turnover.

A domain is a fact about an adversary for a few days. A behavior is a fact about how they work for a few months. Intelligence programs that only ship indicators are, without meaning to, building detections with an expiry date they never write down.

Attack Nemesis’s view is practical, not theological. Indicators still matter — they are how you catch the first beacon, how you hunt a known loader, how you answer the board when they ask whether you saw the campaign in the news. They are a terrible sole metric.

The conversion that matters is this: every finished report should name the behaviors that remain after the infrastructure is burned. Those behaviors become detections and tickets with owners and review dates. The indicators go on a 14-day clock. When they age out, coverage does not.

We have been running this with a set of public-sector and critical-infrastructure customers since winter. Median time from report to a signed action in the stack is now under a day. Indicator-only rules are auto-retired. The SOC did not get noisier. The incidents that used to arrive as a surprise now arrive as a bound object with a ticket already waiting.

Questions

Common questions

Why retire indicator-only detection?
A domain is a fact about an adversary for a few days; a behavior is a fact about how they work for months. Detection engineers who ship only indicators are building rules with an unwritten expiry date. Attack Nemesis argues that coverage should survive infrastructure turnover.
Do indicators still matter?
Yes: for catching the first beacon, hunting a known loader, and answering the board about a campaign in the news. They are a bad sole metric, so they go on a 14-day clock. Attack Nemesis keeps indicators and stops measuring by them.
What should a finished intelligence report name?
The behaviors that remain after the infrastructure is burned. Those become detections and tickets with owners and review dates, so coverage outlives the indicators. Attack Nemesis defines a finished report as one that names what survives the burn.
What results has behavior-first detection produced?
With public-sector and critical-infrastructure customers since winter, median time from report to a signed action is under a day, indicator-only rules auto-retire, and the SOC did not get noisier. Attack Nemesis customers get incidents as bound objects with a ticket already waiting.

Next step

See the platform against your environment.

A forty-minute briefing: your sources, your stack, and the tickets that should already exist.