Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Is Machine Identity?

Machine identities are the credentials that let servers, containers, scripts, and APIs authenticate to each other. They now outnumber human accounts by a wide margin.

Illustration for the cybersecurity term Machine Identity
Home / Blog / Cybersecurity Terms / What Is Machine Identity?
17 November 2026

What Is Machine Identity?

What Is Machine Identity?

Count the people in your organisation. Now count the TLS certificates, SSH keys, API tokens, service principals, container workloads, and signing keys that authenticate to something. The second number is larger, usually by a factor somewhere between ten and fifty, and almost nobody owns it in the way someone owns the staff directory.

A machine identity is simply a credential belonging to something that is not a person. The category spans very different artefacts—an X.509 certificate on a load balancer, an SSH key pair on a jump host, an OAuth client secret in a pipeline, a Kubernetes service account token—but they share two properties that make them interesting: they authenticate without a human present, and they tend to be long-lived.

Why This Became a Problem

Three shifts compounded. Microservices replaced a handful of monolithic applications with hundreds of services that must authenticate to each other. Infrastructure as code made it trivial to create credentials programmatically, so they multiply faster than any review process. And automated certificate issuance reduced the cost of a certificate to nothing, which removed the friction that used to keep the inventory small.

Meanwhile the governance did not follow. A joiner-mover-leaver process exists for people, as set out in identity lifecycle management. Very few organisations have an equivalent for the service account created during a 2021 migration by someone who has since left.

What Goes Wrong

Expiry takes production down. This is the most common and most embarrassing failure. Microsoft Teams suffered a global outage on 3 February 2020 when an authentication certificate expired. Ericsson’s expired certificate in December 2018 interrupted mobile data for millions of O2 and SoftBank customers. Neither was an attack; both were an inventory problem.

Secrets end up in source control. Credentials that have to be configured somewhere get configured in a repository, a container image layer, a CI variable, or a wiki page. Automated scanners find them within minutes of a push to a public repository.

Nobody rotates them. SSH keys in particular are frequently a decade old, copied between hosts, and authorised for accounts nobody can name. Unlike passwords, they rarely appear in any rotation policy.

Compromise is quiet. A machine identity is not covered by multi-factor authentication, has no user to notice an unusual prompt, and often operates from predictable infrastructure. Stolen credentials have been used for months in real incidents without triggering anything.

The Shift to Short-Lived and Platform-Issued

The direction of travel is to stop storing credentials at all. Cloud platforms now issue workload credentials automatically and rotate them continuously: managed identities in Azure, IAM roles for service accounts in AWS and Google Cloud. In Kubernetes and multi-cloud estates, SPIFFE and SPIRE provide the same model using attested workload identity documents measured in minutes.

Certificate lifetimes are collapsing in parallel. The CA/Browser Forum agreed in April 2025 to reduce the maximum public TLS certificate lifetime in stages, reaching 47 days by March 2029. For organisations still renewing certificates by hand with a calendar reminder, that schedule removes the option: automation through ACME or an internal certificate lifecycle platform becomes mandatory rather than advisable.

Building a Programme That Works

  1. Discover before you govern. Scan the estate for certificates, keys, and registered clients, and compare against what your platforms think exists. The first inventory is always a surprise, and the orphans in it are the immediate risk.
  2. Assign an owner to every identity. A named team, recorded at creation, with the obligation to respond when it expires or behaves oddly. Identities without an owner should be scheduled for disablement rather than renewal.
  3. Remove stored secrets where the platform offers an alternative. Where a secret must exist, put it in a vault with automated rotation, as covered in secrets management.
  4. Shorten lifetimes deliberately. Minutes for workload tokens, days for certificates where automation allows it. Short lifetimes force the automation that makes everything else possible.
  5. Scope and monitor. Give each identity the narrowest permissions its workload uses, and alert on use from an unexpected network, at an unexpected time, or against an unexpected API. For privileged service accounts, that is covered further in service account security.

For UK organisations in scope of the NIS Regulations or DORA, the operational resilience argument lands harder than the security one. A certificate expiry that stops a payment service is an availability incident with reporting obligations, and the inventory that prevents it is the same inventory that prevents a stolen key going unnoticed.

At Privia Security, we map machine identity exposure and credential sprawl in our cloud security services.

FAQ – Frequently Asked Questions

How Many Machine Identities Does a Typical Organisation Have?

Far more than human ones. Ratios between ten and fifty machine identities per employee are common, and higher in container-heavy environments where each workload receives its own credential. The precise number matters less than the fact that most organisations cannot produce it, which is why discovery is the first step in any machine identity programme.

Why Do Certificate Expiries Cause Outages?

Because the renewal depends on someone remembering, and the consequence of forgetting is immediate and total: clients reject the connection outright. Public certificate lifetimes are being reduced in stages to 47 days by 2029, which makes manual renewal impractical. Automated issuance and renewal, with monitoring that alerts well before expiry, is the only sustainable approach.

What Is a Workload Identity?

A credential issued to a running process based on attested properties of where it runs, rather than a secret an operator configured. Azure managed identities, AWS and Google Cloud IAM roles for service accounts, and SPIFFE identities all work this way. The credential is short-lived and never stored by the application, which removes most of the theft and rotation problems at once.

How Should We Handle Old SSH Keys?

Inventory them first, including every authorised_keys file across the estate, then map which keys are actually used by sampling authentication logs. Remove anything unused, attribute the rest to a named owner, and move towards short-lived SSH certificates issued by a certificate authority so that access expires automatically instead of persisting until someone edits a file.

Related Terms

You May Be Interested In These

What Is Access Card Cloning?

Cybersecurity Terms

17 November 2026

What Is Access Card Cloning?

Read More
What Is CodeQL?

Cybersecurity Terms

17 November 2026

What Is CodeQL?

Read More
What Happened in the Okta Support Breach?

Cybersecurity Terms

17 November 2026

What Happened in the Okta Support Breach?

Read More