Skip to content
Tech Interview Prep
Technical interview guide

Identity Lifecycle & Provisioning

Automating account creation, access changes, and deprovisioning across a person's entire time at an organization.

Read
45 min
Practice MCQs
25
Interview Q&A
25
Edition
v5
Editorial status
Reviewed

Scope: SCIM RFC 7643/7644, NIST SP 800-53 Rev. 5 and SP 800-63C-4, current CISA, Microsoft Entra, Google Cloud, AWS IAM Identity Center, Okta, OWASP, and OpenID guidance reviewed 2026-09-04.

Curated: · Written: · Reviewed:

25 questions

MCQ-10A SCIM 2.0 service (RFC 7644) holds this User: ``` userName: "nk@example.com" active: true nickName: "Nia" ``` and receives: ``` PATCH /Users/2819c223-7f76-453a-919d-413861904646 { "schemas": ["urn:ietf:params:scim:api:messages:2.0:PatchOp"], "Operations": [ {"op": "replace", "path": "active", "value": false}, {"op": "remove", "path": "nickName"}, {"op": "add", "path": "nickName", "value": "Nina"} ] } ``` What is the resulting resource?
MCQ-11A SCIM 2.0 provisioning service issues bearer tokens to an HR system. The service allows two active tokens per tenant for a 24-hour overlap during rotation. The team rotates tokens every 90 days. The latest rotation happened at 00:00 on Monday: a new token was issued, and the old token was set to expire at 00:00 on Tuesday. The HR system reads its token from a config file only when its process starts; the process has been running since Sunday. The config file was updated with the new token at 00:05 on Monday. The HR system's sync at 02:00 on Tuesday fails with HTTP 401. Logs show the HR system presented the old token. What is the most likely cause?
MCQ-13A provisioning worker pulls events from one shared lifecycle queue in order. One event for a contractor returns HTTP 500 from the target on every attempt; the worker retries it in place, and the new-hire and termination events behind it sit unprocessed for 40 minutes. Which characterization of that event is correct?
MCQ-9A SCIM 2.0 service provider hosts two tenants, each with about 40,000 users. Both tenants' HR systems emit externalId "1042" for their own employee 1042, and the provider stores the value verbatim. Its matcher runs `SELECT * FROM users WHERE external_id = ?` and takes the first row. A PATCH aimed at tenant A's user 1042 changed tenant B's user's displayName instead. Which change makes the PATCH land on the intended user?
MCQ-12A team pushes SCIM events from its IdP to each SaaS app as users change, and treats a 2xx response as proof the app matches the IdP. An audit finds accounts still active for employees who left months ago. Why is periodic reconciliation required to close this gap?
MCQ-25A 4,000-person company provisions 14 SaaS apps from Okta over SCIM 2.0, with Workday as the authoritative HR source. Q3 closed 380 terminations. A post-quarter audit found 27 accounts in App X that could still authenticate 30+ days after termination; the other 13 apps had none. For 353 of the 380 leavers, App X's connector log shows a deprovision PATCH returning HTTP 200; for the other 27, no deprovision call was logged, because those Okta accounts were deactivated by a bulk import that bypassed the connector's event hook. App X's /Users/{id} returns "active": true for those 27, and their userName values match Okta's exactly. The nightly reconciliation job pulls the current list of users active in Okta, looks each one up in every app by userName, and alerts when an app's active flag disagrees; it logged no lookup errors and reported no drift for App X in Q3. Which change to the reconciliation job would have flagged these 27 accounts?
MCQ-24A team must sign off on a SCIM connector from their IdP to a SaaS target before a quarterly access review. Over the last 24 hours: 12,400 create/update events applied, the newest landing 40 s after emission (the sync SLA is 5 minutes); three deletes for terminated employees have been retrying for 6 hours against HTTP 429 responses from the target; the last full reconciliation completed 9 days ago against a 7-day window. Which evidence set is required before the connector can be declared healthy?
MCQ-22At 09:00 HR terminates employee E-1042. The IdP policy computes a diff that removes E-1042 from the prod-admin group and sets active=false. At 09:02 the SCIM connector PATCHes the app. The app returns HTTP 200 with status pending and operationId op-77. The IdP dashboard shows Provisioned. At 09:10 the app's group export still lists E-1042 in prod-admin. Which audit correlation would show why the app still lists E-1042 in prod-admin at 09:10 after the connector's 09:02 HTTP 200?
MCQ-8Your provisioning service creates users with POST /Users against a SCIM 2.0 endpoint. The client timeout is 30 s, but the endpoint's p99 create latency is 45 s during the 09:00 onboarding burst, so about 2% of creates time out with no response. The service retries each timed-out create once, and the identity team finds roughly 40 duplicate accounts after every burst. Each user record carries externalId, the HR employee number, which is stable and unique per person; userName is generated as firstname.lastname and collides for common names. The tenant has SCIM userName uniqueness enforcement turned off. What should the retry do?
MCQ-7An IdP provisions App X over SCIM and writes its immutable user ID into each account's externalId; this deployment of App X matches inbound SCIM users on externalId only, and the username it displays is not a matching key. The offboarding runbook is: disable the IdP user, run a job that walks that user's app assignment and sends active=false to App X, then delete the IdP user after 7 days. An operator reversed the first and last steps for 240 employees, deleting all 240 IdP users on day 0; this IdP's delete is permanent, with no restore or recycle bin. The job ran on day 7, found no assignments, reported 0 users deprovisioned, and exited 0. On day 14 App X's console lists all 240 accounts as enabled, each still holding the externalId written when it was created. Which action will disable the 240 accounts?
MCQ-1A connector writes SCIM users from Workday into Okta and then onto 30 SaaS apps; it locates an existing Okta profile by userName, which Workday builds from the work email, and externalId is never populated. On 3 March a rebrand changes 3,200 of the 4,000 workers' addresses from @oldco.example to @newco.example in one export. employeeId and the active flag are unchanged for everyone, the connector deactivates a profile only when Workday reports the worker inactive, and the 4 March sync ends with zero errors. On 5 March those 3,200 workers reach none of their apps and each has two ACTIVE profiles, the older carrying all 30 app assignments and the newer carrying none; the other 800 workers are unaffected. Which account of the state fits the evidence?
MCQ-14An IAM connector consumes employee lifecycle events from a queue and applies each in arrival order. Every event carries a monotonic `hr_seq` per employee, and the connector persists `last_applied_seq` per identity. For employee E-1042: at 09:00 it applies seq 47 (termination). Account disabled, removed from `prod-admins`. At 09:04 a delayed seq 41 (promotion) arrives. Account enabled, added to `prod-admins`. At 09:05 the connector re-reads the HR record, which shows E-1042 `active` with entitlement `eng-readers` only at `hr_seq` 49. What should the connector do?
MCQ-16A nightly provisioning sync reads `department` from Workday (HR) and from a contractor portal. The attribute is configured with owner=Workday and precedence Workday > contractor. The pipeline's default conflict resolver is last-writer-wins. For 47 users, Workday says `Engineering` (last modified 02:00) and the contractor portal says `Finance` (last modified 03:00). The sync is about to apply changes. What should it do for these 47 users?
MCQ-18A nightly reconciliation job provisions three nested groups to three targets. Engineering holds 40 users directly plus the Backend group; Backend holds 12 users directly plus the OnCall group; OnCall holds 5 users directly, and no user appears in more than one of those direct sets. AD expands nested groups recursively, the SaaS app expands exactly one level, and AWS IAM Identity Center expands none. The job currently compares every target's Engineering member count to 57 and opens drift tickets against the SaaS app and AWS. Which expected count should the job use for each target so that only real drift raises a ticket?
MCQ-2Nine days before a hire starts, a nightly joiner job grants day-one access as the union of three inputs: the reviewed bundle for the hire's HR job code, the reviewed bundle for the hire's HR department, and every entitlement held by the teammate the manager named as the buddy. For this Treasury analyst the job-code bundle carries 9 entitlements and the department bundle 14, and the two bundles share 5. Both bundles are re-approved each quarter against the HR feed. The buddy holds 26 entitlements, 14 of them inside one or both reviewed bundles. How many of the entitlements the hire receives on day one are backed by a reviewed rule for this hire?
MCQ-4An IdP holds each employee's `department` attribute and updates it within five minutes of an HR change. The CRM export API authorizes solely on membership in the `dept-sales` group, which it reads from the IdP's group API. That group is written by a nightly 02:00 job that reads the HR roster and adds every employee to the group for their current department; the job has no removal step. Priya moves from Sales to Engineering, effective Monday 09:00; the roster and the IdP attribute both update by 09:05. On Thursday she can still call the CRM export API. Which change clears her access at the next 02:00 run and keeps it clear for the next mover?
MCQ-5A 1,200-person company grants temporary production access for up to 30 days as an IdP group membership with an expiry timestamp. Removal fans out to three targets: an AWS SSO permission set and an internal admin panel, both SCIM-provisioned and revoked within about 60 seconds of the IdP change, and a Postgres database, where a script runs once a day at 03:00 UTC and drops any role whose IdP grant is no longer active. Last quarter 720 grants expired; the audit found stale Postgres roles for people whose grants had expired, and none in the other two targets. What is the longest a user can keep their Postgres role after their grant expires?
MCQ-20An identity governance platform reconciles a production database once a night. Each run starts at 01:00 UTC, evaluates every grant as of that start, and applies its revocations to the target only when the run finishes, 90 minutes later. A manual exception grant suppresses removal of one entitlement until its own expiry timestamp. Governance policy caps residual access after an exception expires at 4 hours, measured to the moment the target actually loses the entitlement. A contractor's production read role sits under an exception that expires Thursday 22:00 UTC; the handover it covers ends Thursday 20:00 UTC. What is the residual access window, and what does the policy make of it?
MCQ-23The security policy defines leaver control effectiveness as the elapsed time from the authoritative leave event to the last point at which any effective access path is denied. A leaver's timeline: 09:00 HR marks termination in Workday; 09:12 provisioning feed polls HR and enqueues a deprovision job; 09:47 connector picks up the job; 09:48 SSO session revoked; 09:49 refresh token invalidated; 10:35 the SaaS app's local account is disabled by its connector, which syncs every 45 minutes (last sync 09:50, next 10:35); 11:00 next scheduled full sync of all connectors. Which value should be reported for this leaver?
MCQ-17A SaaS wiki is provisioned from Okta over SCIM and users sign in over OIDC. Offboarding sets `active: false` on the wiki account at 10:00. The wiki's OIDC session cookie lasts 12 hours and its personal API tokens expire after 90 days. At 10:05 the departing engineer's laptop still loads the wiki on the existing cookie, and a script still writes pages with their API token. What actually closes the gap?
MCQ-6Meridian Health runs Okta as its IdP for 40 SAML apps. A terminate event from the HR system drives an orchestration job that deactivates the Okta user, sends a SCIM deactivate to each app account, and ends app sessions for the 39 apps that implement SAML single logout. Okta's session lifetime is 12 hours with a 30-minute idle timeout. The in-house Claims Portal implements SSO but not single logout: it mints its own 8-hour session cookie at SAML login, and its admin UI has a per-user terminate-sessions action. Priya's last SSO login was 09:00. Her terminate event was processed at 11:00; the log shows the Okta deactivation at 11:00:12 and the Claims Portal SCIM deactivate returning HTTP 200 at 11:00:40. At 15:00, helpdesk confirms Priya can still open the Claims Portal without signing in. Meridian's offboarding policy requires active access to be closed within 30 minutes of the terminate event. What closes Priya's remaining access and keeps the next leaver inside the policy?
MCQ-3A SaaS app is provisioned over SCIM at offer acceptance, which in this pipeline is 45 days before the start date. The connector's create payload sets `active: true` and attaches the Employee role; the app also keeps its own password login next to SSO, and its password-set link never expires. Nothing re-evaluates the account when HR changes a start date. A hire accepted 45 days out has the start date pushed back 8 weeks. Which exposure does this configuration produce?
MCQ-19A nightly export job runs at 02:00 and authenticates by exchanging a personal OAuth refresh token, issued to engineer A, for a 1-hour access token; it mints a new access token on every run. A is offboarded on Friday: the account is disabled at 17:00, and the IdP revokes all refresh tokens for disabled accounts within 5 minutes. The job's last successful run was Friday 02:00. On Saturday the finance team reports no export file. Platform policy forbids long-lived static keys for scheduled jobs. Which action restores the export while keeping A's account disabled?
MCQ-15Priya left 14 months ago. Her directory account was disabled but not deleted. It still holds eleven group memberships: including `payments-prod-admins`, granted for an on-call rotation that ended three months before she left: plus a personal access token with `repo:write` and a delegated mailbox grant to her former manager. She rejoins next week as a backend engineer on the fraud team, reporting to a new manager. The HR feed emits a new-hire record that reuses her original employee ID and corporate email address. The fraud team's entitlement policy is already defined and grants six groups, none of them production admin roles; the audit requirement is one directory identity per employee ID. What should the provisioning pipeline do with the existing account?
MCQ-21An IGA deployment provisions from Workday, the system of record for employment status, into Active Directory. After a ransomware incident the AD domain was rebuilt from a nine-month-old backup. Since that backup, 1,240 employees have left and 380 role grants have been revoked. The restored domain holds 4,180 enabled accounts, including all 1,240 leavers. The IGA database and its change-detection cursor are intact and current, and the connector runs an incremental sync every 15 minutes. What should the team do before restoring access?