Duty analyst on shift

UK Cyber Defence

Managed SOC · Guide · Detection

Indicators of compromise:examples everysecurity team should know

An indicator of compromise is a piece of evidence that an attacker has been somewhere: an IP address, a file hash, a domain, an odd sign-in, a scheduled task nobody created. This guide explains what indicators of compromise are, gives practical examples across networks, devices, identities, email and the cloud, shows how to validate and prioritise them, sets out how they differ from indicators of attack, and explains what to do when one matches.

Plain EnglishUpdated November 2026Practical examplesFree reputation lookup

01The short answer

An indicator of compromise is an observable trace of an attacker or their activity, used to find, trace and block them.

The definition comes from RFC 9424, the Internet Engineering Task Force's 2023 paper on indicators of compromise, written with authors from the UK's National Cyber Security Centre. It describes them as observable artefacts relating to an attacker or their activities: their tools, their infrastructure, and the techniques they use. In practice an indicator is something you can search your logs for. The address of a server known to control ransomware. The fingerprint of a malicious file. A domain registered last week to imitate your bank. A sign-in from a country nobody in the company has visited. Indicators are cheap to share and quick to match, which makes them valuable; most are also easy for an attacker to change, which is why they are one layer of detection and never the whole of it.

Also called
IOCs, IoCs, or simply indicators
Common examples
IP addresses, domains, URLs, file hashes, registry keys, unusual account activity
Best used for
Finding known threats quickly, and checking whether a published attack touched you
Main weakness
Attackers change most indicators easily; behaviour is harder to hide

14 days

Median time an attacker was inside an organisation before being detected in 2025, up from 11

82%

Of intrusion detections in 2025 involved no malware, only valid accounts and legitimate tools

29 min

Average time for a criminal intruder to move from the first machine to the next

26m

Malicious IP addresses in the threat intelligence our SOC matches against client activity

  1. Median time an attacker was inside an organisation before being detected in 2025, up from 11: Mandiant M-Trends 2026
  2. Of intrusion detections in 2025 involved no malware, only valid accounts and legitimate tools, Average time for a criminal intruder to move from the first machine to the next: CrowdStrike 2026 Global Threat Report
  3. Malicious IP addresses in the threat intelligence our SOC matches against client activity: UK Cyber Defence threat intelligence holdings

What they are

What indicators of compromise are, and why some matter more than others

An indicator of compromise is useful because it can be expressed precisely and searched for automatically. A threat report says that a ransomware group's tool contacted a particular server; anyone with network logs can check in seconds whether any of their machines did the same. That speed is the whole appeal, and it is why advisories from the NCSC and its partners end with a list of indicators.

Not all indicators are equal. In 2013 David Bianco, then at Mandiant, described them as a pyramid of pain: the higher up an indicator sits, the more it costs an attacker to change once defenders detect it. RFC 9424 uses the same pyramid, and adds that indicators at the bottom are more precise but more fragile, while those at the top are more durable but harder to match exactly.

LevelExampleHow easily an attacker changes itHow to use it
File hashesThe SHA-256 fingerprint of a malicious programTrivially: change one byte and the hash changesBlock and search for known malware; confirm a file is the one in a report
IP addressesA command-and-control server, a scanning hostEasily: a new cloud server takes minutesBlock for a limited time; search history for past contact
Domain namesA look-alike domain used for phishingWith some effort: registration, hosting, DNSBlock and alert; watch for new look-alikes of your own name
Network and host artefactsA distinctive URL path, user agent, registry key or file locationIt means retoolingDetection rules in the SIEM and on endpoints
ToolsA particular remote access tool or credential dumperPainfully: new tools must be found or builtBehavioural rules for the tool, whatever its hash
Tactics, techniques and proceduresDisabling backups before encryption, approving an MFA prompt by phoneHardest of all: it means changing how they workBehaviour-based detection mapped to MITRE ATT&CK

The practical lesson is to use the bottom of the pyramid for speed and the top for resilience. A blocked hash stops one copy of one file. A rule that notices backups being deleted at three in the morning stops the next ransomware group too.

The life of an indicator

RFC 9424 describes a lifecycle: an indicator is discovered, assessed, shared, deployed into defences, used to detect, acted upon, and finally retired. The last step is the one most often missed. IP addresses are reassigned, domains expire and change hands, and an indicator that was malicious last year can belong to a legitimate business today. Indicators need an expiry date.

Examples

Indicators of compromise: twelve examples every security team should know

The first five are classic, shareable indicators. The rest describe behaviour, which is where most modern intrusions are found.

01Network

Known malicious IP addresses

Connections to or from servers listed in threat reports: command-and-control, malware download hosts, anonymising services used for sign-ins.

02Network

Malicious and look-alike domains

DNS lookups for domains used by malware, newly registered domains, and names that imitate yours or your suppliers' for phishing.

03Endpoint

File hashes

The SHA-256 of a known malicious program, document or script found on a device or in an email attachment.

04Endpoint

Suspicious files and locations

Executables in temporary or user folders, files with double extensions, and tools such as credential dumpers that have no place on a server.

05Endpoint

Persistence artefacts

New scheduled tasks, services, registry run keys or startup items created outside change control, so that something runs again after a reboot.

06Behaviour

Unusual process behaviour

Word or Outlook launching PowerShell, a browser spawning a command shell, or a backup agent suddenly deleting shadow copies.

07Identity

Unusual account activity

Sign-ins from new countries or anonymising networks, impossible travel, logins at odd hours, and many failures followed by a success.

08Identity

Changes to MFA and privileges

A new multi-factor authentication method nobody expected, a user added to an admin group, or a dormant account suddenly active.

09Email

Mailbox rules and delegation

Inbox rules that forward, delete or hide messages, especially about invoices, and new delegated access to a mailbox.

10Cloud

Cloud application consents

An OAuth application granted access to mail or files, new API keys, or resources created in regions you never use.

11Network

Beaconing and odd traffic

Regular small connections to the same host at fixed intervals, traffic on unusual ports, or large uploads to file-sharing services at night.

12Defence evasion

Log gaps and tampering

Security logs cleared, auditing switched off, endpoint protection disabled or a log source that has gone quiet without explanation.

Where to look

Where indicators of compromise are found

An indicator is only useful if the evidence that would contain it is being collected. These are the main sources, and what each typically reveals.

WhereWhat it recordsIndicators it can reveal
Firewall, proxy and DNS logsEvery connection and name lookupContact with malicious IPs and domains, beaconing, unusual ports, large uploads
Endpoint detection and responseProcesses, files, registry and network activity on each deviceMalicious hashes, suspicious process chains, persistence, credential dumping
Identity provider logsSign-ins, MFA events, role and group changesImpossible travel, anonymised sign-ins, new MFA methods, privilege changes
Email logs and gatewaysSenders, links, attachments, deliveryPhishing infrastructure, malicious attachments by hash, look-alike sender domains
Cloud audit logsAdministrative actions in Microsoft 365, Azure, AWS and GoogleApplication consents, new keys, unusual regions, mass downloads
Server and application logsWeb requests, authentication, errorsExploit attempts, web shells, unusual administrative access

Two gaps are common. Edge devices such as firewalls and VPN gateways often keep little telemetry, and Mandiant's M-Trends 2026 found the longest-hidden intrusions of 2025 were frequently persisting on exactly those devices. And logs that are not kept long enough cannot be searched when an advisory arrives about an attack that began weeks ago.

IOCs and IOAs

Indicators of compromise versus indicators of attack

The two terms are often used together and sometimes confused.

Indicator of compromiseIndicator of attack
The question it answersHas a known threat been here?Is someone trying to do something harmful right now?
What it describesEvidence left behind: an address, a file, a domainBehaviour and intent: a sequence of actions that only makes sense as an attack
When it firesOften after the event, when the evidence is foundDuring the attack, ideally before the damage
ExampleA connection to a server named in an NCSC advisoryAn Office document starting PowerShell, which downloads a file and creates a scheduled task
How durableFragile: easy for the attacker to changeDurable: tied to what the attacker has to do
Typical sourceThreat intelligence feeds and advisoriesDetection rules mapped to MITRE ATT&CK, behavioural analytics

The shift matters because attackers increasingly bring nothing with a hash. CrowdStrike found 82% of its detections in 2025 involved no malware at all: intruders signed in with valid credentials and used tools already on the system. There is no file to fingerprint in a stolen session or a misused remote administration tool, only behaviour. A detection programme needs both: indicators of compromise to catch the known quickly, and indicators of attack to catch the unknown.

Validating

How to validate an indicator of compromise

A match is a question, not an answer. Before acting on an indicator, check five things.

  • The source. Who published it, when, and with what confidence? An NCSC or CISA advisory, a reputable vendor report and an anonymous paste are different things.
  • The context. Shared infrastructure causes false alarms. An IP address belonging to a content delivery network, a cloud provider or a VPN service may serve thousands of legitimate sites as well as one bad one. A hash may belong to a legitimate administration tool that attackers also use.
  • The age. IP addresses are reassigned and domains change hands. An indicator from two years ago may now belong to someone else entirely.
  • What actually happened. Was the connection blocked or completed? Was the file merely downloaded, or did it run? Did the sign-in succeed, and was MFA satisfied?
  • Corroboration. Does anything else agree? A single DNS lookup is weak evidence; the same lookup followed by a new scheduled task and outbound traffic at regular intervals is strong.

Our free Reputation Lookup checks an IP address or domain against public threat feeds and blocklists, and the Phishing Header Analyser shows where an email really came from. For indicators tied to an exploited vulnerability, the CVE Explorer shows whether the flaw is on CISA's list of known exploited vulnerabilities.

Prioritising

How to prioritise indicators of compromise

A busy environment produces many matches, and most are noise. Rank them by what they mean, not by how many there are.

Confidence. How sure is the source that the indicator is malicious, and how specific is it? A hash of a known ransomware binary is specific; a scanning IP that probes the whole internet is not.

Where it matched. The same indicator means more on a domain controller, a finance system or an administrator's laptop than on the guest Wi-Fi.

Whether it succeeded. An attempt blocked at the firewall is a data point. A completed connection, a file that executed, or a successful sign-in is a potential incident.

What it is associated with. Infrastructure used by a ransomware group, or by an initial access broker who sells to one, deserves faster attention than commodity spam.

Relevance. Indicators from campaigns against your sector, your suppliers or the technology you run should rank above generic feeds.

PriorityExampleTypical response
UrgentA hash of known ransomware tooling executed on a serverIsolate the host, start incident response, hunt for the same indicator everywhere
HighA successful sign-in from an anonymising network, followed by a new MFA methodRevoke sessions, reset credentials, check what the account did
MediumA workstation resolved a domain from a recent phishing advisory, no further activityCheck the user and device, block the domain, look for the email
LowRepeated connection attempts from a known scanner, blocked at the edgeRecord and let the block stand

Sources and sharing

Where indicators of compromise come from

Good indicators come from people who have seen the attack. The main sources for UK organisations are:

  • Government advisories. The NCSC, often jointly with partners such as CISA and the FBI, publishes advisories with indicators and detection guidance for significant campaigns. The NCSC's free Early Warning service tells registered UK organisations when their networks appear in threat feeds.
  • Vendor and researcher reports. Security companies publish indicators from the incidents they investigate, usually with the context needed to judge them.
  • Sharing communities. Sector groups and information sharing platforms such as MISP exchange indicators between members, often in the STIX format and over TAXII, the standard ways of describing and moving threat intelligence.
  • Your own incidents. The indicators from an attack on you, or a near miss, are the most relevant of all. Record them and share them with your peers and the NCSC where you can.

A feed of indicators is only as good as what it is matched against. Millions of indicators applied to two weeks of firewall logs, with nobody to read the matches, achieves little.

If an indicator matches

What to do when an indicator of compromise matches

  1. Stage 01

    Confirm it

    Check the source, the context, the age and what actually happened, as above. Most matches end here, recorded with the reason.

  2. Stage 02

    Scope it

    Search for the same indicator across every log you hold, going back as far as your retention allows. One match often has company.

  3. Stage 03

    Contain it

    Isolate an affected device, disable or reset an affected account and revoke its sessions, and block the indicator at the edge and on endpoints.

  4. Stage 04

    Preserve the evidence

    Keep the logs, a copy of the file and, where possible, the memory of the affected machine. Switching off destroys evidence.

  5. Stage 05

    Pivot from it

    Use the confirmed indicator to find related ones: other files from the same folder, other hosts that contacted the same server, other accounts signed in from the same address. Then look for the behaviour behind them.

  6. Stage 06

    Remove the attacker and recover

    Close the route in, remove persistence, rebuild what cannot be trusted and restore from clean backups. Watch closely afterwards for a return.

  7. Stage 07

    Report and share

    Tell your insurer, the ICO within 72 hours if personal data is involved, Report Fraud or Police Scotland where a crime has occurred, and share the indicators with the NCSC and your peers.

In practice

Making indicators of compromise work in practice

Match continuously, and look back. New indicators should be applied to live data and to history. When an advisory names a server used three weeks ago, the useful question is whether anything contacted it three weeks ago. That needs logs kept long enough and a search that can run across all of them.

Expire what goes stale. Give each indicator a lifetime appropriate to its type: short for IP addresses, longer for hashes of known malware. Old indicators cause false alarms and, worse, teach people to ignore them.

Allow-list the known good. Content delivery networks, update services and your own infrastructure belong on a list that indicators cannot override without review.

Automate carefully. Automatic blocking of high-confidence indicators saves time. Automatic blocking of low-confidence ones takes down legitimate services. Keep a person in the decision for anything that would isolate a server or lock an account.

Climb the pyramid. For every indicator that matters, ask what behaviour it represents and write a rule for that behaviour too. Our SOC uses Wazuh as its detection engine, where lists of indicators and behavioural rules sit side by side; the same approach works in any SIEM.

Where SOC365 fits

Where a managed SOC fits

Most organisations do not have people to collect the logs, maintain the feeds, read the matches and decide what they mean at three in the morning. That is the gap a security operations centre fills.

SOC365, our managed SOC, matches client activity against our own threat intelligence holdings, including 26 million malicious IP addresses, 620,000 malicious domains and more than four million indicator hashes, alongside detection rules for behaviour mapped to MITRE ATT&CK. EmilyAI, our AI SOC analyst, reviews every client every hour and runs a structured threat hunt each morning; our analysts check what she raises before it reaches you. When a new advisory is published, we search back through your history for its indicators, not only forward.

Every alert you receive has been looked at by an analyst who can explain it. Clear-cut attacks are blocked automatically at the firewall and on blocklists; isolating a machine or locking an account is decided by an analyst, with your approval or under authority agreed in advance, and incident response is part of the service. Our guide to managed detection and response explains how this fits beside the tools you already have.

Questions

Questions about indicators of compromise

What is an indicator of compromise?

An indicator of compromise, or IOC, is an observable trace of an attacker or their activity, such as an IP address, a domain, a file hash, a registry key or an unusual sign-in. Security teams search for indicators in their logs to find, trace and block known threats.

What are examples of indicators of compromise?

Common examples are connections to known malicious IP addresses or domains, file hashes of known malware, new scheduled tasks or services, Office applications launching PowerShell, sign-ins from unusual countries, unexpected new MFA methods, inbox forwarding rules, cleared security logs and regular beaconing traffic to an unfamiliar host.

What is the difference between an IOC and an IOA?

An indicator of compromise is evidence that a known threat has been present, such as a malicious file hash. An indicator of attack describes behaviour that shows an attack in progress, such as a document launching a script that creates persistence. IOCs are quick to match but easy for attackers to change; IOAs are harder to evade.

What is the pyramid of pain?

A model created by David Bianco in 2013 that ranks indicators by how much it costs an attacker to change them once detected: file hashes, IP addresses and domains at the bottom, then network and host artefacts, tools, and tactics, techniques and procedures at the top. Detecting behaviour causes attackers the most difficulty.

Where do indicators of compromise come from?

From government advisories such as those of the NCSC and CISA, security vendors' incident reports, sharing communities using platforms such as MISP and formats such as STIX, and an organisation's own incidents.

How do you validate an indicator of compromise?

Check who published it and when, whether it belongs to shared infrastructure such as a content delivery network, whether it is still current, whether the activity actually succeeded, and whether other evidence agrees. A single match is a question, not a conclusion.

How long are indicators of compromise useful?

It depends on the type. IP addresses can change hands within days or weeks, domains within months, while hashes of known malware stay accurate for as long as that file exists. Every indicator should be given an expiry date.

What should you do if an indicator of compromise is found?

Confirm the match, search for the same indicator across all your logs, contain affected devices and accounts, preserve evidence, look for related indicators and behaviour, then remove the attacker, recover and report. An incident response team can help with each step.

Can antivirus detect indicators of compromise?

Antivirus detects some file-based indicators, such as known malicious hashes. It does not see network, identity or cloud indicators, and most modern intrusions use valid accounts and legitimate tools rather than malware, so detection needs logs from across the estate.

Do small organisations need to monitor for indicators of compromise?

Yes, but most do not need to build it themselves. Endpoint protection, sign-in alerts and the NCSC's free Early Warning service are a sensible start, and a managed SOC adds the feeds, the retrospective searches and the people to read the matches at any hour.

Indicators of compromise

An indicator only helps if someone is watching when it appears.

Thirty minutes on what you log, how long you keep it, who would see a match at three in the morning and what we would change first. No slides.