IdentityOps Explained in Plain Language, Floor by Floor

Rajesh Kumar

Rajesh Kumar is a leading expert in DevOps, SRE, DevSecOps, and MLOps, providing comprehensive services through his platform, www.rajeshkumar.xyz. With a proven track record in consulting, training, freelancing, and enterprise support, he empowers organizations to adopt modern operational practices and achieve scalable, secure, and efficient IT infrastructures. Rajesh is renowned for his ability to deliver tailored solutions and hands-on expertise across these critical domains.

Imagine a large office building with a thousand staff. At the front desk sits a register of everyone entitled to be inside, and a guard who checks each person’s badge. Now imagine that when people quit, nobody updates the register, that contractors borrow each other’s badges, and that the cleaning robots carry master keys nobody has counted. This is the problem IdentityOps exists to solve, only in software rather than concrete. The explanation below uses plain language and a running office-building picture to cover what the discipline is, how it works and how to tell whether yours is healthy.

The register and the guard

Every company with computers keeps lists of accounts: people, groups, and programs that sign in on their own. IdentityOps is the routine care of those lists and of the proofs people and programs present to get in. The Ops Atlas, which sits on the xOps School site, sums up the focus as knowing who exists and seeing the proof they present. It deliberately leaves out the question of what each account is allowed to do, which it assigns to a related practice, and the question of how extra access is requested and reviewed, which belongs to another.

The Atlas is open about the discipline being new. No official definition exists, and different suppliers describe it differently. Some see it as simply running an identity system after it is installed; others describe a merger of account management, security operations and automation. You can see these competing framings laid out in the Atlas page that compares definitions of IdentityOps.

Three things the discipline watches

1. People arriving, moving and leaving

In the building, this is the badge desk. In software, the trigger is usually a record in the HR system. When someone is hired, the directory creates an account and the applications they need follow. When they change role, some access should be added and some removed. When they leave, everything should close. Atlas readers will know the shorthand JML, for joiners, movers and leavers, and it sits at the middle of the work.

The slip-ups happen at the third stage. A departed person’s account is switched off in the main directory, but the cleanup stops there. Their account in the project tool, the file-sharing service or a small niche application is still active months later. The Atlas names this problem bluntly: disabled is not the same as offboarded.

2. Proof of identity

The guard checks badges. A badge that can be photocopied is a poor badge, and the digital equivalent is a sign-in method that can be tricked. Text-message codes and approve-on-your-phone prompts are common, yet the Atlas classes them as weak because attackers can steal or imitate them. The only options it accepts as phishing resistant are physical security keys following the FIDO2 and WebAuthn standards, and smart cards that use certificates. It adds that passkeys which sync across devices do not reach the very top assurance tier, since their keys can be exported.

Passwords still matter. The advice the Atlas repeats from NIST favours long ones, at least fifteen characters when used alone, with no demand for odd symbols, no scheduled expiry, and a check against lists of known bad passwords. It sounds counterintuitive until you remember that forced changes push people towards predictable patterns.

3. Robots with keys

The cleaning robots in our building stand for what the Atlas calls non-human identities: service accounts, API keys, automation tokens. Nobody schedules an exit interview for a key. They accumulate, and they often have broad reach. The Atlas lists a tidy lifecycle for them. Count them, give each an owner and a purpose, set an expiry and a review interval, prefer credentials that expire fast, replace them regularly, and when one is no longer needed, switch it off, wait to see who complains, then delete it.

Why a switched-off account may still work

Here is the part that surprises non-specialists. When you sign in, many systems hand you a temporary pass, such as a session or a token, so that you do not need to show your badge every minute. If your account is later disabled, the temporary pass may still be valid until it runs out. The Atlas notes that on one major cloud identity service the default pass lasts roughly one to one and a half hours, and that settings built for continuous evaluation can extend that to around a day. Even the call that cancels sessions can take a few minutes, and it does not always reach sessions held inside separate applications.

The building version: you take away a visitor’s badge at the desk, but the door he already walked through stays open behind him. Good practice therefore pairs disabling with an explicit step that cancels live passes, and records the time of each action.

How to know if your identity housekeeping is healthy

Evidence-minded readers will want numbers. The Atlas offers a sample calculation. Take forty leavers in a period. Thirty-six had their directory account closed within the agreed time. That gives a score of ninety percent if you only look at the directory. Yet three of those thirty-six kept a working account in a lesser application, so counting every application drops the score to 82.5 percent. Measure at the far end of the chain, not the first link.

Another example concerns strong sign-in. If 1,190 of 1,400 staff are enrolled in a strong method, coverage looks like eighty-five percent. If only 41,000 of 62,000 sign-ins actually use it, usage is about sixty-six percent, which tells you weaker options are still accepted. Counting usage as well as enrolment catches that gap.

A rhythm you can copy

  • Every day: compare the directory with the HR list.
  • Every week: read a sample of account recoveries.
  • Every month: list machine credentials that are close to expiring.
  • Every quarter: rehearse a leaver, and review which outside providers can vouch for your people.

For quiet accounts, the Atlas proposes a periodic sweep that treats three to six months of silence as the trigger. It leaves out staff on HR suspension, break-glass emergency logins, and machine identities whose owners registered a slower review cycle.

The human side of a technical routine

Rolling out these habits changes how people work, and the cultural side deserves attention. If you want a grounded look at how teams digest new routines, our hands-on advice about steering culture change through DevOps adoption covers it. Staff who are asked to carry a security key may grumble for a week; staff who see a clear reason usually adapt quickly. Explaining the badge-and-register picture helps.

The risk angle matters too. Identity failures sit among the most common root causes in public incidents, and the Atlas notes that the cases it cites defeated process rather than cryptography. A broader look at how reliable delivery and security trade off is in this overview of balancing risk, reliability and secure delivery.

A short example to hold on to

Suppose a design contractor finishes a three-month project on Friday. On Monday morning her directory login is closed, which feels like the end of the story. But she signed in to the shared documents tool on her tablet on Thursday, and that session is still alive. She also created an automation key last month so her slides could sync to a folder, and that key belongs to no one now. Neither gap is dramatic. Both are exactly the sort of residue the routine is designed to catch: cancel the session, find the key’s owner, and note the time against each action so the record is complete.

Plain-language glossary

  • Directory: the master list of accounts.
  • Identity provider: the service that checks sign-ins and issues passes.
  • Single sign-on: one login that opens many applications.
  • Provisioning: creating, changing and removing accounts in applications automatically.
  • Token: a temporary pass issued after sign-in.
  • Non-human identity: an account used by software, not a person.

Takeaway

Keep the register accurate, make the badge hard to forge, count the robots, and measure the exit. Everything else in the field elaborates on those four moves. The scale of the technology can distract from how ordinary the idea is: know who is inside your building, and be able to show how they got in and when they should have left.

Leave a Reply