Top 100 Identity and Access Management (IAM) Engineer Practice MCQs
Every question on one page: open one, choose an answer, and see why at once. Signed in, your latest answer to each question counts, and you can answer again after 24 hours.
Curated: · Written: · Reviewed:
Reviewed 100Review pending 0
Answers to reviewed questions count toward this role and the overall leaderboard; your latest answer to each one is the one that counts.
MCQ-4A service at client-a handles an OIDC callback at 09:35. The ID token's RS256 signature verifies against the kid in the IdP's JWKS, and its claims are iss=https://idp.example.com/t1, aud=client-a, azp=client-a, nonce=n-7f3a, auth_time=09:32, exp=09:42. The service's configured issuer is https://idp.example.com/t1, its stored nonce for this authorization request is n-91c4, and its maximum authentication age is 15 minutes. What should the service do?⌄
MCQ-2A mobile app is a public client using authorization code flow with PKCE against an IdP. A tester registers a second app that claims the same custom URI scheme, captures the authorization code from the redirect, and replays it to the token endpoint 25 times, each time omitting code_verifier. Every attempt returns an access token. What does this establish?⌄
MCQ-5A stateless backend runs the OIDC authorization code flow for a web app. On each login it generates a 32-byte `state`, stores `state -> {code_verifier, redirect_uri}` in Redis with a 5-minute TTL, and redirects the browser to the IdP with `state` and `code_challenge_method=S256`. On callback it looks up the returned `state` in Redis, rejects the request if the key is absent, and exchanges the code with the stored `code_verifier`. An attacker runs the authorization request with a script that does not follow the redirect, captures the callback URL for the attacker's own account, and sends it to a victim. The victim opens it, and the app signs the victim in as the attacker; the attack succeeded in 3 of 3 runs. Which change closes this?⌄
MCQ-6A team's authorization server validates the `redirect_uri` on every authorization request with this function (Python 3.11, `urllib.parse`):
```python
from urllib.parse import urlparse
def accepts(registered, requested):
r, q = urlparse(registered), urlparse(requested)
return q.netloc.endswith(r.netloc) and q.path.startswith(r.path)
```
The client is registered with `https://login.example.com/callback`. Which requested `redirect_uri` does the check accept and deliver the authorization code to a host the client does not control?⌄
MCQ-1Portal is a SaaS app that signs users in with the OAuth 2.0 authorization code flow against a shared corporate IdP. It exchanges the code for an access token, calls the IdP's /userinfo endpoint, and creates a session from the returned sub and email. The IdP issues opaque access tokens, and /userinfo returns no audience and no authentication time. A pen test pasted an access token minted for the Wiki client, a different app registered at the same IdP, into Portal's callback and got a valid session for that user; four such sessions were opened over six days. Which change closes the hole?⌄
MCQ-3A payments API at `https://api.example.com/payments` validates every bearer JWT by checking the signature against the IdP's JWKS, then `iss` (`https://idp.example.com`), `exp`, and `aud` against a single configured value. That value was copied from the mobile app's OIDC client config and left as `mobile-app-7f3a`. Authorization is a lookup of the token's `sub` in the API's own entitlements table; the API never reads `scope`. A tester logs in as `u-4412` through the mobile app and replays the ID token from that login (`aud: "mobile-app-7f3a"`, `sub: "u-4412"`, `exp` 60 minutes out, no `scope` claim) against `GET /payments/accounts`. `u-4412` has `payments:read` in the entitlements table. What does the API return, and why?⌄
MCQ-7A public iOS client sends S256 challenges. Its SDK generates one 32-byte code_verifier at first launch, keeps it in the keychain and reuses it for every authorization request. A second app on the test device has registered the client's custom redirect scheme, so it receives every redirect together with its code, and it also recovered the verifier once from an unencrypted device backup taken on day 1. Over 19 days it replayed 7 intercepted codes at the token endpoint with the public client_id, the hijacked redirect_uri and that day-1 verifier, and all 7 were exchanged for tokens. The only change since is at the authorization server, which now rejects code_challenge_method=plain; nothing in the SDK changed. On day 20 the team reruns the attack, intercepts one code and replays the same day-1 verifier within seconds of the redirect. What happens, and why?⌄
MCQ-8A B2B marketplace lets each seller register an OIDC provider. The app uses one redirect URI, https://app.example.com/callback, for every provider and stores the selected provider in a cookie. A seller registers evil.example. In a drill, the seller starts a login to the corporate provider, then replaces the corporate authorization code in the callback with one from evil.example. The app redeems the evil code at the corporate token endpoint; corporate rejects it. In a second run, the seller sends a corporate code to evil.example's token endpoint; evil.example returns an ID token with iss set to the corporate issuer and sub of a corporate admin. The app verifies the ID token signature against evil.example's JWKS and creates a corporate session. Five corporate accounts were taken over before the mismatch was noticed on day 8. Which change addresses the root cause?⌄
MCQ-16An RP validates tokens by fetching the OIDC discovery document and JWKS on every request and follows HTTP redirects. During a 35-minute window a misconfigured proxy answered the discovery URL with a 302 to a lookalike host; the lookalike document advertised the same issuer string as the real IdP and served its own JWKS. The RP accepted tokens signed by an unrelated key and created 17 sessions. Which single check would have rejected those sessions?⌄
MCQ-15An IdP rotates its token-signing key at 09:00: it publishes the new key in its JWKS, starts signing with it, and removes the old key from the JWKS in the same change. Relying parties cache the JWKS for 24 hours (Cache-Control: max-age=86400), validate strictly against their cached key set without refetching on an unknown key id, and issue access tokens with a 60-minute lifetime. What happens, and what should the team have done?⌄
MCQ-17An internal IdP signs access tokens with RS256 and publishes its signing keys at a JWKS endpoint. Eleven services verify those tokens with a shared in-house helper. A penetration test called the order and ledger services directly on the internal network and got both to accept a token with `alg: none` and a token with `alg: HS256` signed using the IdP's public key bytes as the HMAC secret. Which change should the team make?⌄
MCQ-25A gateway validates 5,000 JWTs/s against an issuer's JWKS, caching the key set for 10 minutes and refetching whenever a token carries a kid absent from the cache. An attacker injects 200 forged tokens per second, each with a random kid, and valid sign-ins begin failing. The JWKS endpoint allows 20 requests per second per client and returns 429 above that. The team requires that a genuine key rotation be picked up within one minute. Which change should the gateway adopt?⌄
MCQ-27Ledgerly's backend creates a session when it receives a bearer token at POST /session: it calls the provider's /userinfo with the token and, if a subject comes back, sets its session cookie. FitTrack, a separate app registered with the same provider, posted the access tokens it holds for 5,400 of Ledgerly's 6,000 users to /session, and all 5,400 produced a Ledgerly session. Which change closes this?⌄
MCQ-26A payments platform validates OIDC ID tokens at its gateway and SAML assertions at a partner-facing service; both receive the same token from the same issuer. Auditors require any validation decision from the last 90 days to be reconstructable: issuer, subject, audience, signing key ID, decision time, pass or fail, and which check failed, and they want a single token event tied across the two services. Security policy forbids retaining bearer credentials in logs of any kind. A scan found 4,200 full ID tokens in the gateway's structured logs at INFO, 380 of them still inside their validity window. Which change satisfies the audit and policy requirements?⌄
MCQ-18An OIDC relying party signs users out by redirecting the browser to the IdP's end_session_endpoint first, and deletes its own session only in the post_logout_redirect_uri handler. On shared kiosks, 1 in 6 sign-outs left a usable app session: the IdP was unreachable or the third-party cookie was blocked, so the redirect never returned and the Redis session record and cookie stayed valid for the remaining 30 minutes of the idle timeout. Which change fixes the failure?⌄
MCQ-24A finance-approvals app federates to the corporate IdP over OIDC. The IdP session lasts 30 minutes, the app's own session cookie lasts 8 hours, and the access token lasts 1 hour; the app never re-checks the IdP after login. An account suspended in the IdP at 10:40 kept approving purchase orders until 18:05, and six approvals in that window had to be reversed by hand. The revocation objective for finance approvals is 15 minutes. Which change meets it?⌄
MCQ-23A SaaS code-review app authenticates users through the company IdP, then issues its own 30-day refresh tokens for its API. SCIM deactivation flips the app's user record to inactive within 2 minutes, and the IdP dashboard shows the user inactive in all 14 connected apps. A departing engineer is deactivated at 09:00 on a Friday; at 09:20 his laptop, holding a refresh token issued 3 days earlier, is still reading private repositories through the app's API. Which explanation and fix is correct?⌄
MCQ-10An admin console maps the OIDC `groups` claim to roles with this function, and `issue_refund` is authorized only when `"billing-admin" in roles`:
```python
ROLE_MAP = {"eng-oncall": "operator", "finance-readers": "reader"}
def roles_from(claims):
return [ROLE_MAP.get(g, g) for g in claims["groups"]]
```
The IdP lets any employee create a group. A contractor's token carries `groups = ["eng-oncall", "billing-admin", "contractors"]` and the contractor opens the refund page.
What does the console do?⌄
MCQ-9A 12,000-seat SaaS app federates to its IdP over OIDC and keys its local accounts on the email claim. The IdP's last 12 months show 318 email changes, 502 display-name changes, and 0 changes to the subject identifier. Last quarter a mailbox was reissued 90 days after the previous holder left; the new joiner's first sign-in matched the old account on email and inherited 3 administrative group memberships plus 4 years of the previous holder's records. Which change removes the inheritance path?⌄
MCQ-22A workforce analytics app creates a local account the first time it sees an ID token for a user, keyed on the `email` claim, and stores the token's `iss` and `sub` on the row for audit. The IdP issues a `sub` once per (issuer, client) and never reassigns it, and omits `email` entirely for users who have no address. A domain migration moved 380 users from @corp-old.example to @corp-new.example and left every `sub` unchanged. Each of the 380 now has two rows carrying the same (iss, sub): the original, created before the migration, with 14 role assignments and 3 report subscriptions, and a duplicate created during the migration with 0 entitlements. Browsers with a session already open kept working until the 8 h session expired, and the next login landed on the duplicate. Which change leaves each of the 380 users in an account that carries the 14 role assignments and 3 report subscriptions, without a per-user ticket, and still holds after the next rename?⌄
MCQ-11A dismissed engineer's IdP account was disabled at 09:12. The wiki's session cookie kept working for 7 hours, and a refresh token at the customer system minted access tokens for 13 days. The runbook's only step is 'disable the account in the IdP.' Which addition closes both gaps?⌄
MCQ-12App A is a SaaS analytics tool that creates accounts just-in-time on first SAML sign-in; no provisioning protocol is configured. The IdP is the source of truth, and the team's deprovisioning SLO is that a leaver's App A account is deactivated within 1 hour of the IdP account being disabled. The IdP can push lifecycle changes to an app within minutes; the team's reconciliation script runs nightly. A review of App A's own account list found 1,286 accounts matching an active IdP identity, 214 with no matching identity, 19 of those holding workspace-admin, and the oldest orphan unowned for 22 months. Disabling a test user in the IdP stopped new sign-ins but left the App A account active. Which change should the team make?⌄
MCQ-13Your SP verifies the XML signature on the SAML Response and then reads the first Assertion element in the document. It serves many tenants through one ACS URL and must keep accepting responses from IdPs that sign only the Assertion as well as ones that sign the whole Response. A pen test produced an admin session by wrapping a signed assertion issued to another tenant with an unsigned admin assertion, and separately by presenting a tenant-B assertion to tenant A's ACS, which the SP accepted because it never compared the Audience. Which change should the review approve?⌄
MCQ-14A service provider runs six nodes behind a load balancer, each keeping its own in-memory set of consumed SAML assertion IDs. The IdP sets NotOnOrAfter five minutes after issuance. A tester captured one assertion from a TLS-terminating proxy log and replayed it against the pool: four of six attempts were accepted, each creating a full session. The team's rule is that a captured assertion must never be accepted twice, even if a shared component is briefly unavailable. Which change meets that rule?⌄
MCQ-21Your SAML service provider has 14 connections, each with its own entity ID and ACS URL; the ACS resolves the connection from the Audience, validates the XML signature, the Audience and a 5-minute NotOnOrAfter window, and requires InResponseTo to match a request ID it stored against the browser's pre-auth cookie and consumed on first use. Product wants AllowUnsolicited set to true on all 14 connections; that flag makes the ACS skip the InResponseTo comparison for the connection. Three customers asked for it: two run IdPs that only speak the IdP-initiated flow and reject any AuthnRequest, and the third has dashboard-tile launch in its contract. The other 11 have never sent an unsolicited response and their admins say they have no plans to. Which change do you approve?⌄
MCQ-20An OpenID Provider issues pairwise subject identifiers. Three relying parties are registered: app.example.com and shop.example.com, both owned by the same company and required to recognize the same user across the two apps, and partner.example.net, a third-party analytics vendor that must not be able to correlate its users with the company's. All three were registered with the same sector_identifier_uri, https://example.com/sector.json, and the partner's logs now show the same sub value the company's own apps receive for the same user. Which change fixes this?⌄
MCQ-28A SAML IdP's signing key was exported to a configuration repository and used to sign 9 forged assertions accepted by 3 relying parties over a 6-hour window. Relying parties cache IdP metadata for 24 hours and do not re-fetch it on signature validation failure. IdP sessions last 12 hours; refresh tokens last 30 days. The team publishes a new key and removes the old key from the metadata 40 minutes after detection. Which containment step is still required?⌄
MCQ-43A team stores each user's TOTP seed in a `totp_seed` column of the `users` table, encrypted with AES-256-GCM under a 32-byte data key kept in a `settings` row of the same database. The database replicates to a read-only analytics replica whose credentials are shared with the BI team. A BI contractor's laptop is compromised and the attacker dumps both the `users` and `settings` tables from the replica. 40,000 users have TOTP enrolled. How many of their second factors can the attacker now forge?⌄
MCQ-44A service verifies 6-digit TOTP codes with 30-second steps. To stop drift complaints the window was widened to ±5 steps, and the verifier keeps no record of used codes. In a red-team run one captured code was relayed and the server accepted it three times over the next 40 seconds, and a scripted guesser found 11 codes valid at any instant. The team ships a fix: window ±1 step, plus a per-user record of the last accepted step that rejects any code at or below it. After the fix, how many codes does the server accept at any instant, and what happens when the same captured code is presented a second time?⌄
MCQ-31A red-team proxy sits in front of the corporate IdP, serves the real login page, and when a user submits a 6-digit TOTP code it replays that code to the IdP inside the 30-second window and captures the resulting session cookie. Over 3 weeks, 88 of 400 targeted accounts were taken over; the median gap between the user typing a code and the attacker's first authenticated request was 9 seconds, and the IdP logged every one of those logins as a successful TOTP verification from the user's own device fingerprint. Which action does the trace justify?⌄
MCQ-19A payments console requires MFA re-authentication before any transfer over $10,000. On success it sets `session.step_up_until = now + 30 min`, and while that flag is set no further challenge is issued. A victim completed a step-up for an $11,000 transfer to payee A. Ninety seconds later an attacker driving the victim's own browser session (an XSS in a support widget, same IP and user-agent) moved 26 transfers of $11,000 each to 26 different payees in 90 seconds. All 26 are logged as authorized. Which change closes the hole?⌄
MCQ-45An attacker with a valid password sent 9 MFA push prompts to one account between 02:10 and 02:14; the user approved the 3rd, and 4 minutes later mailbox forwarding rules were added. The proposed change caps prompts at 5 per hour per account, locks the account after 3 consecutive denials, and keeps one-tap approval. Which statement should the review accept?⌄
MCQ-46An identity platform's MFA verification endpoint is hit by three patterns at once: one code value sprayed across thousands of accounts, many codes tried against a few privileged accounts, and one account probed from many source addresses. A month earlier, an attacker submitting wrong codes tripped a per-account lockout and locked 300 staff out for a weekend; the service desk queue reached 4 hours. Which throttling policy should the team adopt?⌄
MCQ-40An IdP scores each sign-in 0–100 and maps the score to factors: below 40, password only; 40–79, push approval; 80 and above, push plus a FIDO2 key. Over 9 days, an attacker signed in 14 times to 7 internal applications, each time scoring 12–28 and admitted on password alone. The sign-ins came from residential, mobile and corporate-VPN addresses, on the user's managed, patched laptop; the first 4 used a password from a 2023 breach dump, the rest a password the attacker set after a help-desk reset. Which change closes the path?⌄
MCQ-41A retail bank's web app issues a step-up push approval when a customer adds a payee. The approval returns a token with claims {sub, action: add_payee, exp: issued + 15 min}, and the server sets step_up=true on the session for the token's lifetime, accepting any add-payee request while it is set. At 10:00:00 a customer approves adding payee 'Acme Ltd, acct 1234'. At 10:00:40 an attacker holding that customer's stolen session cookie submits add payee 'R. Croft, acct 9999', and the server accepts it. Which change makes the 10:00:40 request fail?⌄
MCQ-42An infostealer lifted an engineer's session cookie from a managed laptop. Forty minutes later the cookie was replayed from a residential IP in another country and reached the admin console; the app accepted it because the cookie is a self-contained bearer token with a 30-day lifetime, and the IdP refresh token lasts 90 days. Which change makes the next stolen cookie fail when replayed from a different client?⌄
MCQ-29An IdP has 20,000 users. 12,000 sign in with a password plus an SMS code delivered to their phone; 6,000 with a password plus a TOTP app on the same phone; 2,000 with a password plus a FIDO2 security key. Every user signs in on their phone, and for the SMS and TOTP groups that phone is also the one that supplies the second factor. A red-team exercise unlocked 40 enrolled phones and signed in to 38 accounts without a further challenge. The team can ship one enrollment-policy change this quarter. Which change should the review approve?⌄
MCQ-30An identity team moves WebAuthn sign-in from `id.example.com` to `sso.example.com`; both sit under the registrable domain `example.com`. The RP ID was set to `id.example.com` when the 3,800 security keys were registered. After the cutover every key fails with `NotAllowedError`, and the failure occurs before any request reaches the server. What restores key-based sign-in for these users?⌄
MCQ-47A 4,000-employee company handles MFA resets on a helpdesk phone queue. The script asks for employee ID, start date and manager's name; median handle time is 6 minutes and the queue clears about 1,200 resets a month. In March a caller answered all three questions about a senior finance analyst from a public profile, had a new authenticator enrolled in 6 minutes, and three payment-detail changes were attempted within the hour; the ticket records no evidence beyond the caller's answers. Which change closes the path that caller used?⌄
MCQ-32An identity team requires FIDO2 security keys for all 42 directory administrators and declares the rollout complete. Three weeks later an attacker runs an adversary-in-the-middle proxy on a lookalike sign-in page. The administrator's key is never touched, yet the attacker replays a captured session cookie and creates a federation trust that stays live for 11 days. Which gap does the incident expose?⌄
MCQ-33A workforce app is rolling out passkeys. Policy requires user verification. After enabling, 12% of sign-ins from a 4,000-user managed Windows fleet fail because the authenticator data has UV=0, though users report using Windows Hello fingerprint. The IdP admin proposes storing fingerprint templates server-side to verify the biometric. What should the team do?⌄
MCQ-34A relying party at rpId example.com runs a WebAuthn registration endpoint. It verifies the attestation signature, requires clientDataJSON.origin to be https://example.com, requires clientDataJSON.type to be webauthn.create, and rejects a credential ID already registered to the same account. The challenge is checked only for being 32 bytes of base64url. An audit finds 3 accounts whose stored registration challenge values never appear in the server's issued-challenge log, and each of those accounts has since authenticated with an authenticator the user does not own. Which change stops this?⌄
MCQ-35A relying party stores signCount per WebAuthn credential. Its fleet is 2,000 credentials: 1,200 hardware security keys whose counters increment on every assertion, and 800 synced passkeys whose counters are permanently 0. Last week it enabled a check that rejects any assertion whose signCount is less than or equal to the stored value. By Monday morning 800 users were locked out, all with synced passkeys, and the team disabled the check for the whole fleet. The 1,200 hardware keys now accept assertions with a stale counter. Which change should the team make before re-enabling the check?⌄
MCQ-36Fourteen administrators enrolled passkeys. The IdP records the WebAuthn backup-eligible (BE) and backup-state (BS) flags at registration, and five of the fourteen admin credentials were registered with BE=true. Six months later an attacker called the help desk, took over one administrator's personal cloud account, and signed in to the admin console from a laptop the company has never issued. The IdP log shows a successful passkey assertion with BE=true and BS=true, from an AAGUID on the approved platform-authenticator list. The attacker held access for five days. Which conclusion and remediation is correct?⌄
MCQ-37An incident review reconstructs a session-theft takeover from the IdP audit log. 09:40. The victim signs in with password + TOTP; the session is issued with amr=[pwd,otp], auth_time=09:40, TTL 12 h. 14:02:10. The attacker replays the stolen session cookie from the same corporate NAT egress IP the victim used. 14:02:40. The attacker changes the account's recovery email to attacker@example.net using that session. 14:03:00. The attacker calls POST /webauthn/register with the session cookie; the IdP accepts because the session's amr contains otp and auth_time 09:40 is inside the 12 h re-auth window. 14:03:20. The attacker deletes the victim's existing passkey. The victim was locked out for three days. Which single change to the enrolment endpoint would have blocked the 14:03:00 step?⌄
MCQ-38An identity platform issues 10 recovery codes per account, each 10 characters drawn uniformly from a 32-symbol alphabet, stored as bcrypt hashes at cost 12. A read-only replica dump covering 40,000 accounts was posted publicly; it contains no plaintext codes. Within 24 hours, attackers signed in to 3 accounts using recovery codes. Which conclusion and next step fit the evidence?⌄
MCQ-39A company issued FIDO2 keys to 4,000 staff and turned off passwords for those accounts. Lost-key recovery is a service-desk call: three knowledge questions (employee ID, manager's name, last four of SSN), then a 72-hour temporary password that can enrol a replacement key. Last quarter the desk took 1,200 lost-key calls, and a red team using a data-broker record and LinkedIn passed the questions in 3 of 10 attempts. Which conclusion and fix are correct?⌄
MCQ-48In March a hospital ward banned personally owned phones at the bedside. The 42 nurses on the ward still need a second factor for the medication-dispensing app, so since then they have shared one ward sign-in whose second-factor code goes to the charge nurse's phone. In 90 days the audit log shows 11,400 dispense approvals, all under the ward account, and a 40 mg morphine discrepancy could not be traced to the nurse who released it. Which change addresses the cause of the unattributable approvals?⌄
MCQ-49A leaver's account was disabled, but a hardware key registered against a second, older account object still authenticated 5 months later, after that object was re-enabled during a mailbox migration. The provider's disable endpoint revokes sessions but leaves registered authenticators intact, the directory holds several account objects per person from past migrations, and one partner IdP holds credentials outside the directory. Which approach should the team adopt?⌄
MCQ-95A file share is granted to the group `finance-readers`. Its owner reports the share reaches 40 people: the direct members of that group. An audit pulls the nesting graph: `finance-readers` has 40 direct members and nests `eu-analysts`; `eu-analysts` has 15 direct members and nests `contractors-eu`; `contractors-eu` has 22 direct members and nests `all-employees`; `all-employees` has 1,800 direct members and nests nothing. No user appears in more than one of these four groups. How many distinct people can reach the share through effective membership?⌄
MCQ-93A login service builds its directory search by concatenating the submitted username into the filter `(uid=<username>)`, takes the first entry the server returns, and binds as that entry's DN with the submitted password. The directory holds 5,600 person entries. During review, an engineer submits `*` as the username: the server returns all 5,600 entries, the service binds as the first one, and the response echoes that DN. Which change closes the injection?⌄
MCQ-94An identity platform syncs users nightly from Active Directory into an application whose account table is keyed on the distinguished name. The sync inserts a row for any source distinguished name not already present and marks any table row whose distinguished name is absent from the source as orphaned. A reorganisation changed the manager attribute for 1,900 users; 1,200 of them also moved between organisational units, which rewrote their distinguished names. Forty of the moved users also changed their account name (sAMAccountName). No objectGUID changed, and the next nightly sync completes without error. What does the application's account table contain afterwards?⌄
MCQ-50A quarterly access review covers 900 roles and 12,000 entitlements. Forty application managers spend 3 weeks each; 92% of entitlements are approved without being opened, and the campaign removes 11 grants. 610 of the 900 roles have exactly one member and hold 4,100 entitlements their holders use daily. HR is authoritative for department and title and updates nightly. Which change cuts the number of review decisions?⌄
MCQ-56A platform's RBAC graph has four inheritance levels. The base role `employee` is inherited by 27 roles; 3,100 users hold a role somewhere in that subtree, and 40 of them hold `employee` directly. A change request adds `billing:read` to `employee` and is approved on the stated impact of "40 users". Nine days later an access review finds `billing:read` on accounts that should not have it. How many users gained `billing:read` from that change?⌄
MCQ-63A policy engine evaluates a rule for production data: `role == 'Engineer' AND classification == 'internal' OR device_managed == true`. The intended meaning is that Engineers may read internal data, or Engineers on managed devices may read it. An audit finds 220 contractors in APAC with managed devices read production data for 17 days. Role assignments and directory claims are correct. What is the cause and the fix?⌄
MCQ-53A storage bucket `audit-logs` is governed by an engine configured with deny-overrides. Four policies are attached:
- P-41: allow `s3:PutObject` on `audit-logs/*` to role `etl-runner`
- P-42: deny `s3:PutObject` on `audit-logs/archive/*` to role `etl-runner`
- P-43: allow `s3:GetObject` on `audit-logs/*` to role `etl-runner`
- P-44: deny `s3:PutObject` on `audit-logs/*` to role `report-reader`
A request from `etl-runner` to put `audit-logs/2024-06-01.json` arrives. Which policies apply to that request, and what does the engine return?⌄
MCQ-52A data lake enforces access with ABAC: an IAM policy grants the `analyst` role `s3:GetObject` when a bucket's `data-classification` tag is `internal`, and grants the `platform-admin` role `s3:DeleteBucket` when the `environment` tag is `sandbox`. Bucket owners hold `s3:PutBucketTagging` on the buckets they own. A nightly job failed on a `restricted` bucket; the owning engineer retagged it `internal`, the job passed, and CloudTrail recorded a routine `PutBucketTagging`. Over the next 6 days 12 buckets were retagged and 260 analysts could read data they had no grant for. Which change removes this class of failure?⌄
MCQ-51A health-data platform decides access with deny-shaped rules, deny if subject.clearance_level < resource.tier, and a request that no rule fires on is allowed. Subject records come from the attribute service, which returns 200 with whatever fields the HR feed gave it. A feed migration dropped clearance_level, so 4,300 of 9,800 subjects became records without the field, and for 90 minutes their tier-4 reads passed the deny rule and reached the log as ordinary allows. The review that followed found 37 rules across six services in the same shape, reading attributes from four different sources; a third of all decisions are machine-to-machine calls. Which fix stops the grants?⌄
MCQ-61A platform team refactors an authorization policy. The text diff is 12 lines and the change ticket says "no access changes". Before deploy, the team replays a recorded sample of 1,300,000 production decision inputs (principal, action, resource) against the old and new policy. The replay shows 41,000 requests that were denied now allowed, 260 that were allowed now denied, and the rest unchanged. All 41,000 newly allowed requests come from 3 service principals that now match a wildcard action on the storage-bucket resource. What should the reviewer do?⌄
MCQ-62An identity platform writes one row per authorization decision, holding timestamp, subject, action, resource and outcome. Policy bundles deploy by overwriting a single mutable policy row, 41 deploys in the past year. The attribute values read during evaluation go to a side table sampled at 100 percent with a 30-day TTL, and decision rows are retained 13 months. The IdP's own audit log records authentication events only. A customer disputes a deny from 7 months ago and the team cannot reconstruct the decision. Which single change makes decisions taken from now on replayable from the log alone?⌄
MCQ-54A multi-tenant SaaS serves GET /v1/reports/{id}. The handler verifies the JWT signature, checks that the token carries reports:read, loads the row by primary key from a shared reports table, and serializes it. Report IDs are sequential integers assigned at creation. In a test, an authenticated user of tenant 41 sent IDs 1000 through 1200 with a valid token: 187 of the 201 responses were 200 OK carrying bodies that belonged to 38 other tenants, and the gateway raised nothing because its alert rule counts only 401 and 403 responses. Which change closes the leak, and what does the test result show?⌄
MCQ-66A team fixes a cross-tenant leak in `POST /v1/docs:batchGet` by checking every submitted ID against the caller before returning it. A pen test submits 60 IDs (41 the caller owns, 12 owned by another tenant, 7 that do not exist) and gets HTTP 200 with the 41 documents plus, for each of the other 19, `{"id": ..., "error": "forbidden"}` for the 12 and `{"id": ..., "error": "not_found"}` for the 7. The pen test files a finding. What is it?⌄
MCQ-60A multi-tenant SaaS exposes 118 tenant-scoped endpoints. Ninety-two declare an {id} path parameter that the test generator reads from the OpenAPI document; the other 26 take the resource identifier in a JSON body or a query string. To close this cycle's access review, the team must show a case for every tenant-scoped endpoint in which a valid session for tenant A supplies tenant B's identifier in each position that endpoint accepts, with zero responses containing B's data. The team seeds A and B in staging, signs in as A, generates one case per declared path parameter, and the pipeline fails the build on any response containing B's field values. The run reports 91 clean cases and one failure: a bulk endpoint returns 30 of B's rows. Which action should the team take before the review closes?⌄
MCQ-58A multi-tenant policy service caches authorization decisions in Redis under the key `authz:{user_id}:{action}` with a 300-second TTL. User 4471 holds `read` on document 88 and nothing else. Over 12 minutes the audit log shows 78 reads of document 91 by that user, all served from cache, and after the grant on document 88 was revoked at 10:00 the cache kept returning allow for it until 10:05. Cached decisions return in 3 ms against 40 ms uncached, and the policy team requires a revoked grant to stop being honored within one second of the write. Which change fixes both failures?⌄
MCQ-59A checkout API asks a remote policy engine for a decision on every request. Its client gives up after 2 s, logs the timeout and allows the request; the engine's own p99 is 180 ms. A 40-minute engine outage last week let 8,200 writes run with no decision, 31 of them from accounts whose entitlement had been revoked that morning. Writes are 4% of requests and reads 96%, and the compliance window for a revocation is 5 minutes. Which change does this evidence support?⌄
MCQ-57A broker grants just-in-time elevation by adding the user to a privileged group at the target for 60 minutes and removing the membership when the timer fires. An auditor checks a user whose elevation the broker marked expired at 12:00. At 12:30 the target still lists the user in the privileged group, and the SSH session the user opened at 11:05 is still accepting commands. The broker's dashboard shows zero active elevations. What does this show, and what should be fixed?⌄
MCQ-65A payments platform issues temporary exceptions to its authorization policy, such as a 10-working-day write grant so a migration can run. The gateway checks the exception row on each request and allows it while now is before expiresAt. An allow is cached for 300 seconds, and every allowed request, including one served from the cache, gets a capability token that downstream services trust for 900 seconds without calling the gateway back. A control now requires that once an exception passes its expiresAt, no request is allowed more than 60 seconds after that time. Which change meets that control?⌄
MCQ-67An export service queues jobs that carry the requester's OAuth access token (1-hour TTL) and calls the records API with it when the job runs. Jobs normally start within 2 minutes, but the month-end backlog pushes p95 wait to 95 minutes, so about 1 in 5 jobs now fails with 401 and is retried by hand. In a separate incident, a job queued 50 minutes before a contractor's offboarding completed after it and exported 8,200 records. Which change addresses both the 401 failures and the offboarding gap?⌄
MCQ-55A payments platform enforces separation of duty dynamically: every finance user holds both `vendor.create` and `payment.approve`, and the approval service refuses a payment when the approver's subject ID equals the vendor's `created_by`. Static separation was rejected because the 5-person finance team all need both entitlements. Four months after go-live, reconciliation finds 9 payments approved by the same person who created the vendor. The vendor portal creates vendors through a shared service account, and every vendor record's `created_by` holds that service account's ID; the human creator appears only in the portal's own event stream. What is the correct fix?⌄
MCQ-64A support engineer transferred to engineering on 3 March. The IdP derives access from the department attribute on the HR record, and the transfer flow grants the new department's 19 birthright entitlements as soon as that attribute flips; nothing removes the old department's 23. Seven of the 23 are held by the engineering role too, including the production deploy entitlement the engineer carries on an on-call rotation that starts 4 March, and the access-request queue quotes a 2-day SLA. Eleven months on, the other 16 are still attached, among them a customer-impersonation right used twice since the transfer, both times from the engineer's own account on tickets they owned. The last three quarterly access reviews closed with no findings. Which change removes the stale grants without breaking the 4 March on-call?⌄
MCQ-68A terminated contractor's browser session still read the finance data warehouse 3 days after the leaver ticket closed. The IdP logged the account disabled 4 minutes after the HR event, and the primary finance app denied the same session within 90 seconds. The team must report one end-to-end access-removal latency figure for the next joiner-mover-leaver cycle that would have surfaced this gap. Which approach is defensible?⌄
MCQ-69You are reviewing a just-in-time privileged access rollout before sign-off. The directory shows 34 standing domain administrators, 12 cloud tenant owners and 46 database owners. The incident that triggered the work: an administrator's workstation was compromised while the account held permanent domain rights, and the attacker created two accounts and a scheduled task within 18 minutes. No approval step slowed them and no elevation record existed afterwards. The approval service has been unavailable for 40 minutes twice this quarter, and the auditors want a single number they can re-derive on demand. Which statement should the review accept?⌄
MCQ-72A quarterly review of a PAM deployment found 61 elevations to the `prod-db-admin` role over 7 months that were approved by the requester's own on-call teammate. Both engineers are permanent members of `dba-approvers`, the group that approves this role, because the same directory rule that grants `prod-db-admin` also populates `dba-approvers`. Elevations last 4 hours. Which change removes the peers' ability to approve each other while still allowing legitimate elevations?⌄
MCQ-71A 900-person company gives each of its 40 domain admins one account, used for email, browsing and the admin portal. After an incident, forensics show the attacker phished one admin, then 6 minutes later reached the admin portal from that admin's own managed laptop on the corporate network, reusing the browser session that was already signed in. No password was entered and no MFA prompt fired. The account's mailbox also held the portal's password-reset messages. Which change removes the path the attacker used?⌄
MCQ-70A nightly maintenance job runs on an EC2 instance whose instance profile carries AdministratorAccess. The job reboots the staging instances the scheduler hands it; an Auto Scaling group replaces those instances daily, so their IDs change. The security review requires the profile to carry only what the job needs. Which change meets that requirement?⌄
MCQ-73A JIT broker checks out a shared root password for 4 hours. The vault rotates the password after each checkout, but the rotation job has been failing silently on 40 of 380 accounts. A password checked out six weeks ago still authenticates to one of those 40 accounts. The team shortens the checkout TTL to 20 minutes. What happens to that six-week-old password?⌄
MCQ-74Post-incident review at a payments company: someone dropped the ledger table on the production primary at 02:14. Vault's audit log shows four engineers read the shared deploy password in the hour beforehand, so nobody can be tied to the DROP. Every connection to the primary authenticates as deploy, whether it comes from the bastion or from a laptop over the VPN. Passwords rotate every 90 days, and a copy a leaver took was still working 41 days after their last day. Auditors now require that each privileged database action be attributable to one named human and that no credential stay usable without a live approval. Which change satisfies both requirements?⌄
MCQ-75A legacy ERP admin account, erpadmin, is shared by 4 engineers and 2 vendor contractors. The ERP logs the account name, source IP and session start/stop times; the jump host logs the named user per session but keeps only 7 days of history. Last month the ERP log shows 38 erpadmin sessions: 31 from the jump host, 7 from a vendor VPN range. A config change at 02:14 stopped a production line for 9 hours; the matching ERP session ran 02:11–02:19 from the jump host. The incident is 30 days old. What should the team do?⌄
MCQ-76An audit of a PAM deployment finds that a database password typed at a prompt was captured in a session recording. The archive was readable by an operations group of 22 people, and the credential was still valid 3 months later when the exposure was found. The auditor will re-test by sampling 500 recordings and requiring zero hits for known secret patterns before the archive is trusted. The team has two weeks and cannot change the PAM vendor. Which remediation is defensible?⌄
MCQ-77Acme's only identity provider is Okta, and its two break-glass accounts are Okta-federated with Okta Verify. During a 4-hour Okta outage both break-glass sign-ins failed exactly as ordinary users' did, and access returned only when Okta support restored the tenant. The review requires a break-glass path that still works when all external identity traffic is blocked. Which change should it require?⌄
MCQ-78A PAM platform governs 1,400 privileged accounts and keeps its own audit log with 30-day retention. Three engineers hold platform-administrator rights. Last quarter one of them disabled session recording for a single target, used a shared root credential for 26 minutes, then re-enabled recording; the platform's log for that window was later trimmed, and the gap surfaced only 4 months later when the SIEM's copy showed 26 minutes with no session events. Which change actually closes this gap?⌄
MCQ-79An IAM team runs a PAM broker that issues time-bound local-administrator grants to 610 Windows servers and records each grant in its own database; the broker is the only sanctioned path for granting local admin. A quarterly reconciliation enumerates privileged membership directly on the targets and finds 572 entries: 402 match grants the broker still lists as active, 27 match grants the broker has already expired and dropped from its active set, and 143 have no broker record at all. Which conclusion and remediation does the reconciliation support?⌄
MCQ-80A payments service authenticates to an object store with workload identity federation. The store's trust policy accepts a token only when `aud` is `objectstore` and `sub` is `system:serviceaccount:payments:api`, and it is evaluated on every request. Calls from the `payments` namespace succeed. Calls from the `payments-batch` namespace, which runs the same image under service account `payments-batch:worker` and projects tokens with `audiences: [objectstore]`, fail with AccessDenied. What is the cause?⌄
MCQ-81A privileged-access broker sits in front of 312 hosts. It mints a short-lived credential from the host's stored key and injects it after an approval names one target. Grants record the target name exactly as the approver typed it; at session time the handler takes the target from the request, matches it to the grant name case-insensitively, and injects the credential held for the host it resolved. The inventory lists `ci-07` on the build VLAN and `CI-07` on the payments VLAN as separate hosts, registered by different teams. On Tuesday an operator whose grant named `ci-07` requested `CI-07`; the match passed, the payments credential was injected, and the shell landed on the payments host. The operator already held an active approval for that payments host. Which change removes this class of failure?⌄
MCQ-82Northwind's support engineers need privileged access to the payments production database. Northwind has no identity provider, rotates its on-call engineers weekly without notice, and the support contract runs 12 months. The company already runs a PAM gateway with session recording and a sponsorship workflow in its IdP. Audit requires that every privileged session be attributable to a named human within 30 days. Which change meets the requirement?⌄
MCQ-83A quarterly recertification campaign covers 12,000 entitlements. Reviewers see group names only, and the UI's default action is 'Approve all'. The campaign closes at 98% approved with a median of 9 seconds per item. A post-hoc sample of 50 approvals finds 31 that the approver cannot justify. For the next campaign the team changes exactly one thing: no response now means removal after a 14-day grace period, while the UI still shows group names only. That campaign closes at 96% approved with a median of 11 seconds per item, and a 50-item sample finds 29 approvals the approver cannot justify. What does the pilot establish?⌄
MCQ-84An AWS organization has 340 accounts, 12 of them production. A quarterly review of the 12 production accounts' CloudTrail shows zero root sign-ins, and the team reports root rarity as satisfied. An incident review then found that in 3 of the 12 accounts the account trail had been stopped and an IAM user with AdministratorAccess created in the same session; those events appeared only in the organization trail delivered to the management account, and were noticed 6 days later when a second account's logs were correlated. Which change to the rarity check would have surfaced those 3 sessions while they were happening?⌄
MCQ-85A platform team onboards engineers by cloning the entitlements of the most recent hire with the same job code. The birthright policy for job code SRE-2 (US, full-time) lists 12 entitlements. Four hires in that job code over the past year ended up with 14, 15, 15 and 17 entitlements; each hire's own approved requests added 2, 1, 0 and 2 respectively. The most recent hire's account is now under access review. How many of its 17 entitlements should the review require to be removed or individually justified?⌄
MCQ-86An identity platform creates accounts from the HR feed 14 days before each start date, enabled, with a temporary password mailed to the personal address on file. The feed carries a start date that is not cleared when an offer is withdrawn; withdrawals arrive as a separate record the integration ignores. Over six months, 40 offers were withdrawn, 12 of those accounts were still enabled on the would-be start date, and one was used to read a shared drive 3 weeks after the withdrawal. Which change makes a withdrawn-offer account unable to authenticate?⌄
MCQ-87A 4,000-person company runs mover changes as two independent jobs. The add job grants the new role's birthright groups and is triggered by the access ticket. The remove job strips the old birthright groups and is triggered by a nightly reconciliation that enqueues a mover only when the HR record's department field changes. Last year's audit of 340 movers found 51 still holding their old birthright group 30 days after the move, and 6 still holding it after 6 months. The team proposes reversing the order so removal runs first. What will that change alone most likely do?⌄
MCQ-88A payments company grants project access through IdP groups. A nightly SCIM sync makes each group's membership match the project's roster in the project tracker: it adds and removes members, and never deletes the group. Each project record carries an end date, and the roster is kept after the project ends as a record of who worked on it. Entitlement review is quarterly and samples 10% of groups. The settlement database's audit standard says access must end on the project's end date and no later.
The Orion ledger-migration project ended 2024-04-30. On 2024-12-01 an audit found 47 members in the Orion project group: 3 had left the company and their accounts were disabled by HR on their last day, 12 had moved on to other projects, and 32 were still employed with no current need for the data. Orion was still marked Active in the tracker and its PM had moved to another team. Which change ends this access on 2024-04-30?⌄
MCQ-89An engineer resigns at 10:00; the identity team disables the Okta account and revokes Okta sessions at 10:02. At 10:40 the SOC sees the same user still pulling code from GitHub Enterprise and still holding a VPN tunnel. Okta logs show no authentication after 10:02. GitHub Enterprise is federated by SAML and issues its own 24-hour session cookie; the VPN validates SAML only when a tunnel is established, and tunnels last 12 hours. Which statement explains the 10:40 access and names the step the kill chain is missing?⌄
MCQ-90A SCIM connector provisions workers from an HR system into an IdP. Each create sends userName = the work email and externalId = a UUID the connector mints fresh for every HTTP attempt. The target enforces uniqueness on externalId and answers a duplicate with 409; userName is not unique-constrained. The target's edge proxy aborts any request at 30 s with a 504: a vendor-fixed limit the connector team cannot change. Over one 40-minute window, 300 creates got a 504: 220 had already committed at the target, 80 had not. The connector retried each 504 once and every retry returned 201. 220 workers ended up with two accounts and their group memberships split across the pair. Which change to the create path stops this recurring?⌄
MCQ-91A SCIM connector pulls identity changes from one queue and applies each batch in a single transaction. A new hire's displayName contains a character the downstream HR system rejects, so the batch fails and the connector retries the whole batch every 5 minutes. Within an hour 1,100 changes are queued behind it, including 3 terminations that completed 26 hours late. Which change fixes the failure mode?⌄
MCQ-92A staff engineer left 14 months ago. Their IdP account was disabled, not deleted, and no group memberships were removed; 4 of the 23 memberships grant payroll approval and production deploy. They are rehired into a support role whose birthright set is 9 groups, 2 of which they already held, and the reactivation provisions that birthright set. The IdP's SSO session lifetime is 30 days. The team reactivates the account. What access does the rehire effectively hold on day one, and what is the correct remediation?⌄
MCQ-96A Windows file service authorizes SMB access from the group SIDs in the Kerberos service ticket's PAC. Alice opens the finance share at 09:00, and the KDC issues her a service ticket whose PAC lists Finance-Approvers. At 10:00 she is removed from Finance-Approvers, but she keeps opening files on the share until 19:00 and every open succeeds. The domain's maximum service ticket lifetime is 10 hours and its maximum TGT lifetime is 10 hours. Which change makes the share refuse her within 15 minutes of the 10:00 removal, without waiting for her to log off?⌄
MCQ-97An internal HR portal runs on WEB01 under the service account svc-web and queries SQL01 directly. Users sign in with Kerberos (Integrated Windows Authentication), and SQL01 enforces row-level security keyed to the connecting user. WEB01 is currently trusted for unconstrained delegation. A red-team exercise on WEB01 used that trust to obtain a service ticket to a file server as one of the signed-in domain admins. The portal must keep showing each user only their own records. Which change should the team make?⌄
MCQ-98A departing engineer's AD account is disabled at 09:15. At 09:40 the engineer's still-running laptop, connected to VPN since 08:00, opens \\fileserver\finance and reads a file. The domain's Kerberos policy sets a 10-hour TGT lifetime, a 7-hour service-ticket lifetime, and a 7-day renewal window. Which statement explains the access and the control that would have stopped it?⌄
MCQ-99A team raised the LDAP diagnostic logging level on the PDC emulator only, ran it for a week, and read the unsigned-bind events: 41,000 binds from 26 appliance addresses and 8,300 from 12 in-house script addresses. They treated that as the complete inventory and set LDAPServerIntegrity to 2 on all six domain controllers. Within the hour, 9 appliances that had never appeared in the log lost directory authentication. What explains the miss?⌄
MCQ-100A single-domain forest has four domain controllers. An attacker held Domain Admin from Thursday 14:00 until containment on Monday 09:00. The team restored DC01 from a Thursday 02:00 system-state backup onto an isolated network, seized the FSMO roles, and reset the krbtgt password once at Monday 11:00. Kerberos policy sets the maximum ticket lifetime to 4 hours and the maximum renewal to 7 days. At Monday 14:00, before any second krbtgt reset, the team reconnects DC01 to the production network. A user then presents a TGT that was issued at Monday 10:30 and signed with the original krbtgt key. What does the domain do with that TGT?⌄