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-1A team reports 85% detection coverage across 40 techniques, based on its rule inventory. A red-team exercise then runs all 40 techniques. Results: 11 produced an alert with enough context to act on; 18 had no telemetry enabled at all; 8 had telemetry but no rule matched; 3 produced an alert that lacked the context needed to act. Which report and next step is correct?⌄
MCQ-19An EDR console reports 100% coverage of the 1,200 hosts it knows about. Network discovery sees 1,380 devices on the same address space and the directory holds 1,310 computer accounts. A red-team intrusion that began on a contractor laptop ran for three weeks with no agent telemetry. Which action closes the coverage finding?⌄
MCQ-26A quarterly patch report built from the agent console shows 1,176 of 1,200 hosts patched: 98%. A reconciliation against network discovery, DHCP leases and the cloud API finds 1,380 hosts, and a scan of the 180 hosts the console does not list finds none of them at the current patch level. What figure and finding should the report carry?⌄
MCQ-3Endpoint telemetry from 340 of 1,200 hosts stopped reaching the SIEM after an agent upgrade; no rule fired for six weeks and the gap was found by hand. The agents are vendor-managed and cannot be modified or supplemented this quarter, and total ingest is dominated by a proxy feed that hides endpoint drops in the aggregate. The team wants a gap detected within the hour. Which control should they add?⌄
MCQ-32A SOC's detection engineering team has 40 ATT&CK techniques in scope for its EDR and SIEM. Last quarter a purple team exercise ran all 40 with the detection engineers in the room: 29 techniques produced no alert, 8 produced an alert the on-call analyst escalated, and 3 produced an alert the analyst closed as benign. The 29 no-alert gaps were remediated and each re-tested; all 29 now alert, and the 3 triage misses were fixed by adding asset context to the rule. This quarter the team can fund one exercise. Which exercise should it fund, and what does that exercise add?⌄
MCQ-98A SOC's SIEM has 412 enabled detection rules. Six months ago the SIEM was upgraded. Last week a red-team exercise ran a credential-dumping technique on an in-scope host that was reporting events normally, and the rule that should have caught it did not fire. The rule-health dashboard reports all 412 rules as enabled and error-free, and this rule is expected to fire at least weekly on this estate. Which change would have surfaced the broken rule before the exercise?⌄
MCQ-91An enterprise has 4,000 endpoints. 3,800 run managed EDR and forward logs; 200 contractor laptops do neither. The detection dashboard reports 92% ATT&CK technique coverage: 46 of 50 techniques have a mapped log source in the SIEM. In last quarter's red-team exercise, an operator used a cached domain admin credential on a contractor laptop to authenticate to the backup server and read 1.2 TB of backups; no alert fired. Detection engineering has 40 backlog items for next quarter. What should the lead do first?⌄
MCQ-40A payments team tracks median time-to-detect (TTD) over the incidents it discovered. Q1: 40 incidents, 8 first reported by an external party, median TTD 14 days. Q2: 40 incidents, 18 first reported by an external party, median TTD 2 days. Which reading of these numbers is correct?⌄
MCQ-33A SOC's quarterly review compares Q1 with Q4. Alerts handled rose from 4,200 to 5,900. Median time to contain rose from 4 h to 11 h. The share of incidents the SOC found itself fell from 70% to 55%. The SOC lead presents the 40% rise in alerts handled as evidence the detection programme improved. Which conclusion does the data support?⌄
MCQ-2A SOC reviews one quarter of detection-rule performance:
| Rule | Fires/week | True positives | Precision | Median triage |
| --- | --- | --- | --- | --- |
| A | 200 | 3 | 1.5% | 40 s |
| B | 12 | 7 | 58% | 9 min |
| C | 2 | 2 | 100% | 22 min |
Last month Rule A fired on a real intrusion; the on-call analyst closed it in 40 seconds as another false positive, and the intrusion ran for 9 more days. Which change best restores the team's ability to catch the next intrusion?⌄
MCQ-5A detection PR adds this rule for a technique that ends with a payload running: alert when a process writes a file with an `.exe` extension under `%TEMP%` or `%APPDATA%`, and a process whose image path is that same written path starts within the same process tree. The red team's four variants all end with the payload running: V1 writes `%TEMP%\p.exe` and starts it directly; V2 writes `%TEMP%\p.exe` and starts it with `cmd.exe /c start`; V3 writes `%TEMP%\p.dll` and runs it with `rundll32.exe %TEMP%\p.dll,Entry`; V4 writes `%TEMP%\p.exe` and registers a scheduled task that starts it, so the payload runs as a child of `svchost.exe`. Which variants does the rule fire on?⌄
MCQ-17A red-team exercise recompiled the credential-dumping tool from source and renamed the binary; the detection keyed on the tool's file hash never fired. The team has one sprint before the next exercise. Which approach should it adopt?⌄
MCQ-31A 30-day hunt tests one hypothesis: an adversary on a compromised host dumps LSASS credentials with a signed LOLBin (`rundll32 comsvcs.dll, MiniDump`) instead of dropping a tool. The query joins Sysmon EID 10 (ProcessAccess to `lsass.exe`) with Sysmon EID 1 (process creation, command line). It returns 412 EID 10 events across 37 hosts: 400 have the signed backup agent as parent, identical hash on all 37 hosts; 12 have `powershell.exe` as parent, all on host-42. On host-42, 9 of the 12 have a matching EID 1 with command line `rundll32.exe C:\Windows\System32\comsvcs.dll, MiniDump <pid> C:\Windows\Temp\lsass.dmp full`; the other 3 have no EID 1 record because EID 1 was disabled on host-42 from day 12 to day 19. What is the correct outcome to record for this hunt?⌄
MCQ-79A SOC's proxy logs show 1,200 managed endpoints. One workstation has made an outbound TLS connection to 203.0.113.77 every 60 seconds, with inter-arrival times between 58 and 62 seconds, for 11 weeks. Each request is 1.2 KB and each response is 48 KB. The destination is not on any threat feed, the IDS has no signature for it, and the volume-anomaly detector is set to 5 MB/hour per host. Payload inspection is not possible. Which detection step would have surfaced this connection as beaconing before the destination was known to be malicious?⌄
MCQ-78A breach is being scoped. Two questions must be answered: which internal hosts connected to 203.0.113.44, and which files were read from the file server. The perimeter sensor keeps a 4-hour rolling packet capture and the intrusion began 11 hours ago. Traffic is TLS 1.3 end-to-end with no interception. The file server runs an EDR agent that records file reads, and flow logs are retained for 30 days. Which evidence plan answers both questions?⌄
MCQ-30A SOC with 90-day log retention pushed 180,000 indicators from a commercial feed to its perimeter blocklist over one quarter; the feed's median indicator lifetime is 2.5 days. The quarter's only intrusion used C2 domains registered 6 hours before first contact, was found 11 days after the fact, and matched a TTP report the team had received but never turned into a detection. Which change addresses the failure?⌄
MCQ-82A SOC receives about 320 endpoint alerts a day, and its seven analysts have roughly 2,500 minutes of combined triage capacity. Median time to close an alert is 25 minutes, and a two-week time study puts 22 of those minutes on looking up the process tree, its parent chain, the user account and the host's activity over the previous 24 hours. All four fields already sit in the EDR index that the detection rule queries, and the rule's logic has not changed in a quarter. Which change cuts the median time to triage a single alert the most?⌄
MCQ-97A SIEM team must bring daily ingest to 2.8 TB/day or less. It currently ingests 7.3 TB/day: web access 5.0 TB/day (appears in 12% of investigations), authentication 0.3 TB/day (94%), DNS 0.8 TB/day (71%), and endpoint process 1.2 TB/day (88%). Web access logs must stay searchable in the SIEM for live triage; the compliance archive is write-only and retrieval takes 24 hours. Which plan should the team adopt?⌄
MCQ-92A 3,000-host estate runs a detection pipeline that produced 62,400 alerts last quarter, with 75 confirmed intrusions. Alongside it sit four decoy service accounts that no legitimate process authenticates as; last quarter they produced 5 authentication alerts, all traced to attacker activity, and the SOC treats that stream as its only zero-triage signal. Compliance will not exempt the decoys: from next quarter its scanner authenticates as each of the four accounts every 30 minutes, from the same subnet as the rest of the service traffic, and the authentication logs record the account and the result and nothing else. Assuming attacker use continues at last quarter's rate, what should the SOC expect from the decoy stream next quarter?⌄
MCQ-93A retailer moved 60 of its 180 workloads to a cloud provider; those 60 run on managed services where the host agent cannot be installed. The on-premises detection set is 180 rules exercising 34 ATT&CK techniques, 70% of them keyed on process execution. Three weeks after cutover those rules fire zero times, the cloud estate has 4 rules and 0 techniques exercised, the provider's audit log delivers 1.2M events/day with no rules attached to it, and the first cloud intrusion was reported by the provider. Which action should the detection engineering lead take?⌄
MCQ-70A vendor's EDR datasheet claims 100% detection of unknown attacks, measured in their lab. Your team runs the 40 techniques from your threat model against the product in your estate: at the default setting it detects 6 of 40 with 12 false positives a day, and at the sensitive setting it detects 22 of 40 with 200 false positives a day. Your two-analyst SOC can triage about 25 alerts a day before genuine incidents start queuing. What should the team do?⌄
MCQ-69A 5-analyst SOC is reviewing a detection platform before purchase. The licence is £45k/year and deployment is quoted at 3 weeks. A 30-day pilot on 500 endpoints produced 2,400 alerts and 48 confirmed true positives; tuning during the pilot consumed 80 engineer-hours. The vendor quotes 10 h/week of steady-state operating effort, while the team's own measurement from the pilot's final week is 26 h/week. Each analyst has 4 h/week of unallocated time. Which conclusion should the review accept?⌄
MCQ-35A SOC's SOAR playbook auto-disables accounts whenever the 'impossible travel' rule fires. The rule fires 400 times a week and its measured precision is 5%. The team wants to keep automating. Which change is justified by these numbers?⌄
MCQ-84A SOC triages 1,200 alerts a quarter. Escalation requires the analyst to judge an alert 'probably malicious'. Last quarter 240 alerts were escalated and 216 were closed as benign, each costing 45 minutes of IR time; the IR team's ceiling is the 180 hours a quarter it already spends. A review of 300 of the 960 closed alerts found 4 that should have been escalated, in four different patterns; one was initial access that ran 9 days. Leadership has named the 9-day dwell the failure to fix. Which change should the team make?⌄
MCQ-81A detection pipeline reads a Kafka topic into a parser, pushes parsed events onto an unbounded in-process queue, and drains that queue into a single correlation engine that keeps per-host session state and sustains 25k events/s. Baseline load is 20k events/s. A volumetric attack holds the topic at 90k events/s for four hours. Parsed events occupy about 700 bytes each, the pod's memory limit is 4 GiB, the topic retains 200 GB, and offsets are committed as soon as events are read from the topic. Ninety seconds into the attack the pod is OOM-killed and restarts, and the events that were in the queue are gone. Twelve high-value source classes account for 10k events/s at peak; the rest is verbose telemetry. The IR team must afterwards be able to say which source classes were captured in full during the window. Which change should the team make?⌄
MCQ-85A SOC is reviewing a detection change that makes a one-line note the default handover for active investigations. In a 12-case pilot, all cases had an open lead at handover. Five one-line handovers repeated 2 h, 4 h, 3 h, 5 h and 2.5 h of work, and 4 of the 5 lost the lead. Four checked-and-ruled-out case records without live contact repeated 18 min, 22 min, 30 min and 15 min of work, with 0 lost leads. Three of those records plus a live handover with the outgoing analyst available repeated 4 min, 8 min and 0 min, with 0 lost leads. Which conclusion follows from the pilot?⌄
MCQ-83Two of the security team's last five incidents ran long. In the first, the on-call engineer, who had never worked on that system, spent 40 minutes working out the correct command to isolate the host. In the second, the responder identified the right action but could not establish who was allowed to approve it. The team is rewriting its runbooks. Which format should they adopt?⌄
MCQ-39A 400-person SaaS company reviews a breach that ran for 9 days. The timeline: a phishing email at 09:12 on day 0, a VPN login with the stolen credential at 09:40, an EDR alert on the workstation at 09:41 that the console auto-closed as a known admin tool, lateral movement to the build server on day 2, 40 GB exfiltrated over six days, and discovery on day 9 when a customer reported a leaked API key. The review closed with one action item: phishing-awareness training for all staff, owner HR, due next quarter. What should the review produce instead?⌄
MCQ-38A payments team runs its incident-response plan every quarter. Q1 was a walkthrough: the on-call lead read the 12 steps aloud, the team agreed each was clear, and one gap was logged. A stale pager number. Q2 ran the same scenario hands-on against a staging clone of production with the on-call rotation: 4 of the 12 steps completed unaided, and the other 8 stalled. 3 on break-glass credentials nobody in the rotation held, 2 on a runbook command pointing at a host decommissioned in 2023, and 3 on no one knowing which team owned the ledger database. The next audit is one quarter away. Which action follows from these results?⌄
MCQ-37A 900-person logistics firm is hit by ransomware at 03:00 on a Saturday: 30 TB across 400 servers is encrypted and the crew demands $1.2M within 48 hours. The last full backup is 26 hours old. Measured restore throughput on this hardware is 1.5 TB/hour, and verifying the restored data adds 4 hours. The cyber-insurance policy covers ransom up to $1M and requires the insurer to be notified before any payment; counsel has advised that any payment must clear sanctions screening first. The runbook names no decision owner, and under the firm's delegation of authority only the CEO can authorise a payment. The CISO is on a flight until 09:00. The incident commander asks the on-call engineer what to do. Which action fits this situation?⌄
MCQ-36A 12-VM production estate backs up nightly at 02:00 to a domain-joined NAS with 30-day snapshot retention; the backup service account is a Domain Admin because the agent needs VSS access on every host. A separate weekly job exports the same data at Sunday 03:00 to an S3 bucket with Object Lock in compliance mode (90-day retention), written by an IAM role with no AD trust. The most recent export completed successfully at Sunday 03:00. Ransomware detonates Wednesday 03:00; the operator uses the backup service account to delete every NAS snapshot and the NAS volume. What is the team's actual recovery position?⌄
MCQ-34At 09:12 EDR flags a rename burst on four workstations with `vssadmin delete shadows /all /quiet` in the command line; the encryptor is still running, and the shared backup service account behind it has authenticated to 60 hosts in the last 48 hours. EDR network isolation is available and no legal hold is in place yet. What is the correct first-hour remediation?⌄
MCQ-51A bank's incident policy defines the start of the regulatory notification clock as the moment the incident is declared, and the regulator allows 72 hours from that point. On a Friday, an analyst sees anomalous outbound traffic at 11:00 and opens a ticket; the incident is declared at 15:00 the same day and escalated to legal at 09:00 on Monday. The team's draft notification is ready at 16:00 on Monday. Which statement is correct?⌄
MCQ-45Hypothetical: your team is signing a contract with a SaaS vendor that will process customer PII. Your regulator requires you to notify affected individuals within 72 hours of the breach. The vendor's discovery is the earliest point at which you can become aware of a breach. Your incident-response runbook budgets 24 hours from the vendor's notification to complete assessment and file the regulator notification. The vendor's draft clause sets its notification deadline from its own discovery and proposes a 30-day outer bound. What is the longest notification period the contract can allow and still let you meet the regulatory deadline?⌄
MCQ-52You are incident commander for a checkout outage. At 14:05 a deploy raised payment authorization errors to 30% of attempts. You posted at 14:12: 'Investigating elevated checkout errors.' At 14:40 there is still no second update. Support has 40 tickets; two enterprise customers have emailed their account managers asking whether to fail over to their backup payment provider. Engineering found the root cause at 14:35 and expects a fix deployed by 15:10. What should you do next?⌄
MCQ-7During a checkout outage, the incident log records: 01:12 the 5xx alert fires; 01:14 the on-call engineer acknowledges; 01:31 the engineer identifies a bad deploy as the cause and posts in #incidents asking who can approve a rollback; 02:40 the service owner, who had been asleep, grants rollback authority; 02:52 the rollback is applied and checkout recovers. About 40,000 payment attempts failed during the outage. Which change to the response process would most reduce the impact of the next incident of this shape?⌄
MCQ-6A payments API alerts on the vendor default p99 > 200 ms for 5 minutes. Over 30 days it fires 90 times: 78 between 23:00 and 01:00, when the nightly settlement batch holds p99 near 450 ms, and 12 spread across the rest of the day. Outside the batch, p99 sits between 120 and 180 ms. Nine of the 12 daytime fires preceded customer-visible incidents, and those incidents peaked at 380 ms; three of the nine never raised the error rate. Which alerting policy should the team adopt?⌄
MCQ-50A SOC ingests corporate-proxy logs into its SIEM with the full request URL and a 400-day retention. The only stated detection use case is exfiltration of customer records, and the team's investigation window for that use case is 30 days. A privacy review finds the URL field carries employee search queries and account IDs, and that retention exceeds the stated purpose. Which change resolves the finding while keeping the exfiltration detection working?⌄
MCQ-49Over 18 months a 40-engineer platform team logged nine insider incidents: five accidental deletions by engineers holding standing production admin rights, two deploys run with a stale credential, one leaver whose production credentials stayed valid 19 days after their last day, and one deliberate exfiltration by an engineer with standing access to the customer database. The SIEM generates about 4,000 alerts a day, 3% are triaged, mean time from alert to triage is six days, and the monitoring programme costs $610k a year. Which change should the review fund first?⌄
MCQ-11A 4,000-employee company has run quarterly phishing simulations for three years; the click rate fell from 18% to 12% and has not moved since. Last month an attacker sent 60 emails to the 40-person finance team. Eleven users clicked, eleven typed their credentials into a lookalike SSO page, and the attacker relayed each login to the real IdP in real time, satisfying the push MFA prompt. Two wire transfers totalling $310,000 left before the fraud team noticed. Which change would most reduce the loss from the next campaign built the same way?⌄
MCQ-23A supplier's mailbox is compromised. A $480,000 invoice arrives from the supplier's genuine domain: SPF, DKIM and DMARC all pass and the display name matches the account manager on file: asking that the remittance account be changed to a new bank. The AP clerk calls the phone number in the email signature; the person who answers confirms the change, and the payment goes out. Which single change to the payment process would have stopped this payment?⌄
MCQ-48A laptop was taken from a desk at lunchtime while its owner was away; the screen was unlocked. The person who took it used the open session to reach internal file shares and the VPN. Full-disk encryption was enabled on the laptop. Which change addresses the cause?⌄
MCQ-87A six-person security engineering team has 120 engineer-hours a month that are not already committed, and one quarter's budget. Four proposals compete for it, and last quarter a phishing email to a contractor mailbox led to stolen credentials and 90 hours of incident response.
A: phishing-resistant MFA (FIDO2 keys) for all 900 staff. Three weeks of setup with two engineers, then 2 h/month for enrolment and resets; it removes the credential-phishing class.
B: an EDR detection content pack with weekly tuning. Two weeks of setup with one engineer, then 40 h/month of analyst triage; it detects three attacker behaviour classes (credential theft, SMB lateral movement, pre-encryption ransomware staging) and removes none of them.
C: a weekly authenticated external scan. One week of setup with one engineer, then 12 h/month; it reports misconfiguration instances across four exposure classes and removes none of them.
D: network segmentation blocking SMB lateral movement between workstation VLANs. Six weeks of setup with one engineer, then 30 h/month of rule review as endpoints move; while the rules hold, the lateral-movement class is gone.
Which proposal do you fund?⌄
MCQ-88A retailer's post-incident review traced a path from an internet-facing VPN appliance, unpatched since a vendor advisory 11 months earlier, into the payment database. For 3 quarters the security team used its 15-minute board slot to present the open-finding register: 41 findings, CVSS-scored, counted by severity. The minutes record no funding decision in any of those quarters. The next meeting has room for one item, and the same two remediations are still unfunded: upgrade the appliance for 180,000 dollars in 6 weeks, or segment the payment network for 420,000 dollars in a quarter. What should the team change for that meeting?⌄
MCQ-68A 130-person SaaS company employs one security engineer. She also handles access reviews and incident response, which leaves 8 hours a week for new control work; the quarter gives her 104 hours. Five initiatives are on the board with the estimates the team already agreed:
| Initiative | Setup | Ongoing |
| --- | --- | --- |
| Phishing-resistant MFA (FIDO2) on the nine SaaS apps holding customer data | 40 h | 1 h/week |
| Immutable backups plus quarterly restore drills for the primary database | 24 h | 1 h/week |
| Closing the internet-facing patching gap (14 hosts with known critical CVEs) | 16 h | 2 h/week |
| Self-hosted SIEM with 12 detections | 220 h | 18 h/week |
| Third-party risk questionnaire refresh, which the largest customer asks for by name | 60 h | 6 h/week |
The two-year incident log holds 6 credential-stuffing or password-reuse incidents against customer-data apps, 2 exploits of unpatched internet-facing hosts, and 1 failed restore test. Which plan should she commit to for this quarter?⌄
MCQ-100A 400-person SaaS company has 2 engineer-years of security budget for the next 12 months, approved on one condition: by month 12 the company must be able to show that last year's incident would not succeed again. That incident: a reused password on the internet-facing VPN, which was fully patched and had no MFA; 9 days of lateral movement; the backup server encrypted along with production; 11 days to recover. The breached account's owner had completed both prior phishing trainings and had the company password manager installed. Today MFA covers 40% of accounts but not the VPN or the 14 domain-admin accounts, the last successful restore test was 14 months ago, the scanner reports 61 internet-facing services (9 unowned), and the SIEM ingests 3 of 11 log sources. Which allocation of the 2 engineer-years is defensible?⌄
MCQ-64A payments team's threat model for a new card-vault service ran for 90 minutes after the design was frozen. It produced 40 threats, all closed in the tracker within two weeks, and no change to the design. Seven months later an incident crossed from the public API tier into the vault subnet over an unauthenticated internal call: a boundary the model's own diagram had drawn but never asked a question about. The next service freezes its design in one week. Which change to how the model is run is most likely to prevent a repeat?⌄
MCQ-10An engineer's SSO password is phished at 09:12 and the attacker signs in. At 09:20 the attacker mints API key K1 in the CI system and clones repo A; at 09:30 it mints K2 from the same session. The team resets the password at 09:40 and does nothing else. At 10:05 the attacker clones repo B with K2. SSO sessions last 12 hours; API keys never expire. Which action set ends the attacker's access?⌄
MCQ-12Northwind runs a 4,200-account mail tenant. MFA is enforced on the web and mobile clients. Over 12 months the identity team logged 61 account takeovers: 44 arrived over the legacy IMAP/SMTP endpoint, which still accepts a username and password; 9 came from session-cookie phishing on the web client; 5 from helpdesk resets; 3 from shared service accounts. Each of the 44 IMAP logins succeeded on the first attempt with a valid password that had been set 14–26 months earlier and entered a public breach corpus only after it was set. The password policy is 90-day rotation with an 8-character minimum and three of four character classes, costing 340 support hours a year. Which single change removes the largest share of the 61 takeovers?⌄
MCQ-13A red-team relay test against your admin login flow produced these results: push approvals relayed in 5 of 5 attempts, one-time codes in 4 of 5, origin-bound passkeys in 0 of 5. In every successful relay the attacker reused the victim's browser session, the login came from the victim's usual ASN and device fingerprint, and the event was recorded as MFA-compliant; the attacker then held the admin session for 3 hours. The detection team proposes one change. Which change should the review accept?⌄
MCQ-14A flat /16 holds 1,400 workstations and 60 servers, all imaged from one build that sets the same local administrator password on every host. An attacker has SYSTEM on three workstations and has dumped the local SAM from one of them; the dumped credential authenticates to all 1,460 hosts. The team can make one change this week. Which change leaves the attacker able to authenticate to the fewest hosts, and how many?⌄
MCQ-16A backup agent runs on 240 servers under a single service account that is a member of Domain Admins. The account's password sits in a config file on every one of the 240 hosts and has not been rotated in 5 years. The team replaces it with 20 per-set accounts: the 240 hosts are split into 20 sets of 12, each set's account is scoped to its own 12 hosts and is not a Domain Admin, and the credential is stored only on those 12. An attacker compromises one server, copies that set's credential, and is not detected. How many of the 240 servers can the attacker authenticate to with that credential?⌄
MCQ-15A domain administrator's credential was recovered from a compromised helpdesk workstation and used to reach a domain controller 20 minutes later. The team's threat model now treats every lower-tier host as potentially attacker-controlled at the moment an administrator authenticates to it. Which tiering design holds under that threat model?⌄
MCQ-95A 12,000-user Microsoft Entra tenant has 180 privileged role holders. Two are break-glass accounts with no compliant device; they are the documented emergency path. The other 178 split into Group A (120 admins, all with compliant devices), Group B (19 admins, all with compliant devices), and Group C (39 admins, none with a compliant device). A new conditional access policy CA002 will require a compliant device for all privileged role holders. The team must enforce CA002 for the 178 non-break-glass privileged role holders before a 72-hour change freeze, and must keep at least one break-glass account able to sign in. Group C admins have been issued compliant devices and can enrol them within 24 hours. Which rollout plan prevents a lockout?⌄
MCQ-96Entra ID tenant: 4,100 employees, 900 contractors. One Conditional Access policy, 'Require compliant device', is scoped to the Employees group and the Customer Portal app. The Contractors group has a separate policy requiring MFA only. Access tokens live 60 minutes, refresh tokens 90 days, continuous access evaluation is off. A contractor's personal laptop picked up an infostealer. Three hours after the contractor signed in, the attacker replayed the refresh token from a residential proxy in the same country and called the export API, which authorizes on the customers.read scope alone: pulling 48,000 records in six minutes. The export is invoked by a headless sync client on the laptop using the signed-in user's token, so no interactive prompt appears in that flow. No EDR runs on unmanaged devices. Which single change would have blocked this replay?⌄
MCQ-94A 4,900-account company runs no VPN: its 260 internal services sit behind a shared front proxy that admits any request carrying a token the company IdP signed and that has not expired. Access tokens last 1 hour, refresh tokens last 14 days, and a refresh call carries no device credential. An infostealer on a contractor's managed laptop copied the browser's SSO cookies, and there was no sign-in event on the account after day 0. The review's logs show the account's requests arriving from a hosting provider's netblock for 9 days, roughly 216 hourly refreshes, reaching 31 services, until the project ended and the account was disabled. Nothing alerted: the SOC's IdP rules cover failed sign-ins only. Which single change would have ended the 9-day access with no analyst in the loop, and would end the next stolen cookie the same way?⌄
MCQ-8At 02:14 an EDR alert fires on web-07, a PCI-scoped Ubuntu 22.04 host: a process named kworker (PID 4412) holds an outbound TLS session to 203.0.113.47:443 and reconnects within 60 s of every kill. `ls -l /proc/4412/exe` returns `/memfd:kworker (deleted)`, and a systemd unit sysupdate.service restarts it at boot. The runbook says 'power the host off to stop the beacon.' What should the responder do first?⌄
MCQ-9At 09:12 an EDR alert on host APP-14 shows a scheduled task named WinUpdateCheck, created at 02:40, that launches a DLL from C:\ProgramData\ and beacons to 203.0.113.44 every 60 seconds. The task was created with the domain service account svc_backup, which logs on to 9 of the estate's 400 hosts. EDR covers 260 hosts; the other 140 run a legacy agent that logs process creation but not network connections. The IR lead plans to reimage APP-14 at 10:00. What should the team do before that reimage?⌄
MCQ-18A 40-node Kubernetes cluster's worker nodes are all built from the image base-node-v7. node-17 starts beaconing to a known C2 address. The responder kills the process, deletes the payload, and removes the systemd unit that started it. node-17 is drained and reimaged from base-node-v7, rejoining the pool at 09:19; the beacon resumes at 09:24. A node provisioned from base-node-v7 at 08:40, which has not yet been scheduled any pods, carries the same unit. Which action stops the beacon across the cluster?⌄
MCQ-20A 900-endpoint estate is 60% developer workstations that rebuild and patch weekly, 30% servers and 10% fixed-function kiosks. An estate-wide allowlisting rollout last year drew roughly 400 exception requests in its first week and was switched off after three weeks. The next assessment is six weeks away, two engineers own endpoint security, and the EDR is licensed for detection only. Which rollout plan is defensible?⌄
MCQ-21A change hardens macro and script execution across 1,400 Windows endpoints, all enrolled in Intune with the Office ADMX templates applied; today the only control is a warning banner users click through. Twelve users run an unsigned vendor-supplied macro workbook that the vendor will not re-sign before the change window, and the workbook is required for a regulated monthly filing. Incident data shows macro documents arriving by email, web download and chat. Which control set should the review approve?⌄
MCQ-46OrderBridge is a 2013-era order-management host that terminates the EDI feed for 14 trading partners. The vendor is gone: the host cannot be patched, and installing host-based security software stops the EDI parser from starting. It has to keep running for 9 months, stay reachable from the ops jump host, and write a nightly batch to one finance share. Today it accepts connections from any address in the corporate /16 (65,536 addresses), outbound traffic is unrestricted, its shared service account is a local administrator on 60 other hosts, and the SIEM receives no events from it. Which plan bounds what an attacker gets from this host while the feed keeps running?⌄
MCQ-47A plant's safety instrumented system (SIS) runs on its own switch, separate from the basic process control system, and sends its trip data to the plant historian over a one-way link. The SIS has 14 logic solvers; the vendor certifies each one only in its shipped configuration and states that it must receive only its own control protocol and no other traffic. The site's safety case credits the SIS as independent of the BPCS, and the proof test interval is 12 months with the last test 11 months ago. A review proposes three changes: install an endpoint agent on the SIS logic solvers so the SOC can see them, run a weekly authenticated vulnerability scan against the SIS, and add a firewall rule letting the corporate SOC reach the SIS over the process network. Which set of changes can be approved?⌄
MCQ-43A payments team runs a self-hosted GitLab instance. Engineers push directly to `main`, there is no merge-request gate, and CI clones every branch on push. Pushes come from local clones, CI jobs, and a shared build box, and the repo's hook template is not installed everywhere. A nightly post-push scan found a static AWS access key 40 minutes after it was committed; CI had already cloned the branch and two forks existed. Rotating that key takes days because a vendor holds a copy. Which control should the team adopt?⌄
MCQ-44The CI gate for a service blocks a build only when the OSV advisory database reports a vulnerability in a dependency. Last quarter the build pulled `jsonparse` v2.4.0: 0 advisories, 1 maintainer, 2M weekly downloads, no release in 14 months. Two months later the maintainer account was transferred and v2.4.1 shipped a credential-stealer; OSV published an advisory 9 days after that. The same service also depends on `semvertools`: 0 advisories, 1 maintainer, no ownership change, last release 3 years ago, stable and needing no upgrade. The team must change the gate so this does not recur. Which change is defensible?⌄
MCQ-42A payments team runs Semgrep as a blocking CI gate on every pull request. The full ruleset produces 1,240 findings a week across 60 engineers; the triage log shows 4% are resolved or closed as false positives, and the gate was disabled on day 11 after three teams routed around it. A six-week pilot that gates only on the 35 rules that have historically produced exploitable findings produces 48 findings a week and a 71% resolution rate. Which change should the team ship?⌄
MCQ-22Company.com publishes `v=spf1 include:_spf.google.com -all`, DKIM with `d=company.com`, and DMARC `p=reject`. Marketing starts sending a newsletter through a new ESP. The ESP sets the envelope sender to `bounce.esp-mail.net`, signs with `d=esp-mail.net`, and puts `news@company.com` in the From: header. Gmail rejects every newsletter, and the DMARC aggregate reports show each message failing DMARC: the SPF result is recorded against `esp-mail.net` and the DKIM `d=` is `esp-mail.net`. What is the correct fix?⌄
MCQ-53A service issues a session token as base64(JSON{user, role}) and trusts the role field when the token is returned. The change under review adds SHA-256 hashing of the token before it is written to the audit log. The requirement is that a client can neither read nor alter its own role claim. Which statement should the review accept?⌄
MCQ-54A backup service encrypts each 64 KiB chunk with AES-256-GCM under a key rotated once a year. The 96-bit nonce is a 32-bit random prefix drawn once at process start, followed by a 64-bit counter incremented per chunk. The service runs in a VM that is snapshotted nightly; on restore the process resumes with the same prefix and the counter value held in the snapshot. After one restore, the service re-encrypted 40,000 chunks that had already been encrypted before the snapshot, under the same key and the same nonce sequence. An attacker captured both ciphertext sets. Which capability does the attacker gain?⌄
MCQ-55A Go 1.22 service calls `https://ledger.internal:8443` every 30 seconds with a `tls.Config` that sets `InsecureSkipVerify: true` and `ServerName: "ledger.internal"`. The flag was added during an incident to get past an expired internal CA certificate; the certificate was renewed two weeks ago but the flag stayed. A tester on the same VLAN ARP-spoofs the service's gateway and answers with a self-signed certificate whose SAN is `ledger.internal`. What does the tester end up with?⌄
MCQ-56An auth service signs 15-minute access tokens with a single RSA key. Verifiers fetch the JWKS from a well-known URL and cache it for 10 minutes. Ops rotates by replacing the signing key and the published JWKS in one deploy, publishing only the new key. Within a minute, fresh logins fail with 'invalid signature' while sessions established before the deploy keep working. What explains the pattern?⌄
MCQ-57A single Python 3.11 process serves a members' portal. At import it runs random.seed(int(time.time())), and each session ID is built by calling random.choice('0123456789abcdef') sixteen times; nothing else in the process touches that Random instance. Process records show the running instance started between 04:12:00 and 04:12:59 UTC. Sessions stay valid indefinitely, the portal accepts any correctly formed session cookie with no attempt limit, and an outsider can register as many accounts as they like and read the cookies issued to them. The attacker is after the session cookie of a member who logged in earlier that day. What is the attacker's position?⌄
MCQ-58A tester opens `https://app.example.com/?sid=8f2c41d0` in a kiosk browser. The app adopts the query value as the session ID and keeps it after the victim signs in. The tester then sends `Cookie: sid=8f2c41d0` from a laptop that egresses through the same NAT address as the kiosk, 30 seconds after the victim's login, and reads the victim's account page. The session cookie is already `HttpOnly; Secure; SameSite=Lax`, and idle sessions expire after 20 minutes. Which single change breaks this attack?⌄
MCQ-59A web app authenticates with a session cookie set explicitly as `SameSite=Lax`. It exposes `GET /account/email?new=<address>`, which changes the account email and checks neither an anti-forgery token nor the Origin header; the `POST` form of the same change does check a token. An attacker page on another origin embeds `<img src="https://app.example/account/email?new=attacker@evil.test">` and also carries a link to that same URL. A logged-in victim loads the attacker page, then clicks the link. What happens?⌄
MCQ-61A Java service deserialises the base64-encoded Java object in the `session` cookie on `POST /cart`, and the same endpoint also accepts `application/json` bodies for the same operations. A detection rule matches the base64 prefix `rO0AB` in the `session` cookie. Over 30 days it fires 41,000 times, every one from the company's own mobile app. A pentest then achieves remote code execution with a 1.9 KB `application/json` body to `POST /cart`, and the rule does not fire. Which diagnosis fits?⌄
MCQ-62A Django 4.2 app stores user avatars in /srv/app/media, and nginx serves that directory at https://app.example.com/media/. The same origin as the app, whose session cookie is scoped to app.example.com. Filenames are kept as the client sends them, and the allowlist accepts .png, .jpg, .gif and .svg; 38% of the 12,400 stored avatars are SVG. Responses carry the type nginx derives from the extension, so an SVG goes out as image/svg+xml with no Content-Disposition. A pentester uploaded logo.svg containing <svg xmlns="http://www.w3.org/2000/svg"><script>fetch('/api/me')</script></svg>, sent a logged-in user the direct link, and the script ran with that user's session. Avatars must keep rendering inline in <img> tags on app pages, and SVG uploads must keep working. Which change removes the execution?⌄
MCQ-63A SaaS team is fixing its login endpoint after a pentest. It returns 'no such user' for an unknown username and 'wrong password' for a known one, and the wrong-password path takes about 40 ms longer because it runs bcrypt. Each IP is already limited to 10 attempts per minute. Support wants to keep telling merchants why a login failed. Which change should ship?⌄
MCQ-74A log-ingestion worker validates each line with the pattern `^(\w+\s?)*$` before parsing. Normal lines match in under 1 ms, but a single 35-character line, 34 `a` characters followed by `!`, holds one worker at 100% CPU for about 30 seconds, and events queue up behind it. The worker has no regex timeout and runs on the JVM's default engine. Which change removes the stall?⌄
MCQ-77A detection-enrichment worker consumes JSON log lines from a Kafka topic at about 4,000 records/s. For each record it reads the `cmd` field and runs `subprocess.run(f"grep -c {cmd} /var/log/audit.log", shell=True)`. During a red-team exercise a crafted log line ran `id` on the worker and its output appeared in the worker's log; the worker runs as a non-root service account on a host that holds a read-only SIEM API token. Which change should the review require before approving the fix?⌄
MCQ-41An order service has 22 HTTP handlers and one framework request interceptor that every handler passes through. Authorisation is hand-rolled: a handler is expected to call canView(user, orderId) or canEdit(user, orderId) itself, and a code audit found 8 of the 22 do not - 4 read routes (GET /orders/{id}) and 4 write routes (refund, cancel, address change, reissue). Those 8 handlers account for all 8 staging findings and both production findings recorded in the last 24 months.
| Stage found | Findings | Median rework per finding |
| --- | --- | --- |
| design review | 5 | 0.5 engineer-days |
| code review | 20 | 1 engineer-day |
| staging DAST | 8 | 10 engineer-days |
| production | 2 | 30 engineer-days |
Total: 162.5 engineer-days. The staging scan authenticates as two personas and exercises 12 of the 22 routes; refund, cancel, address change and reissue are not among them. The 8 omissions are spread across 6 of the 9 backend engineers, and those same engineers wrote handlers that do check ownership correctly. Two engineers have one sprint, 10 engineer-days, with no other security work, and the goal is to remove the largest share of that rework. Which action should they take?⌄
MCQ-60A link-preview service fetches user-supplied URLs. It runs on EC2 with an instance role attached, and partner URLs redirect up to three times, so redirects have to be followed. Which control set is defensible?⌄
MCQ-66A payments API behind an ALB is guarded by six controls: a WAF rule that blocks /admin unless the source IP is in allowlist.yaml; an app middleware that rejects /admin unless the caller IP is in the same allowlist.yaml; a security group that admits 443 only from 10.20.0.0/16; Okta push MFA enforced by the SSO proxy in front of the app; parameterized SQL in the payments service; and a VPN posture check that also calls the Okta API. The Terraform module net-policy renders allowlist.yaml and the security group, and mounts the file into both the WAF config and the app container. Which pair is genuinely independent: no single failure removes both?⌄
MCQ-67A retail platform runs a quarterly control-testing cycle over 48 detective controls. Seven of them parse the identity provider's audit-log schema, which has changed four times in the last 12 months; each time, those seven stopped firing and the quarterly test caught the break about nine weeks later. The last full cycle passed on 4 April, the schema changed again on 12 May, and the next cycle is due on 4 July. The team's risk register requires any detection control to be re-verified within 7 days of a dependency change. Testing one control by hand takes 20 minutes, and the team has 8 engineer-hours a week for control testing. Which testing plan meets the risk register and fits the budget?⌄
MCQ-71A log-correlation job must join 12 million events against 400,000 threat indicators within a 10-minute batch window. Events arrive as a stream and cannot be materialized or sorted without exceeding the 64 MB executor memory budget. The current nested-loop join has run 9 hours on 32 cores without finishing. Which change will let the job meet the window?⌄
MCQ-72A detection counts failed SSH logins per source IP over a 10-minute sliding window and fires when an IP exceeds 50 failures. The current code, on each event, scans the full retained log for the last 10 minutes and counts that IP's failures. At peak the service ingests 4,000 events/s, so each event scans about 2.4 million events, and the detector's lag grows by roughly 40 minutes every hour; alerts fire after a brute-force burst has ended. Which change removes the lag?⌄
MCQ-76An analyst-authored Python enrichment step sits in the streaming path of a detection pipeline. At peak the pipeline ingests 4,000 events/s; the per-event enrichment loop sustains 900 events/s, so the in-memory queue grows and events are dropped. The detection rules are written and maintained in Python by the analyst team and must stay in Python for the next quarter, and the pipeline must retain every event. Which change closes the gap?⌄
MCQ-80An authorization service caches each decision in Redis with a 10-minute TTL, and every API node keeps its own in-process copy of decisions it has read for 2 minutes; a node that misses its in-process cache reads the Redis entry and copies it locally with a fresh 2-minute TTL. Offboarding policy promises that a revoked permission stops working within 60 seconds, and the authorization service already runs at 70% of its provisioned capacity at the current miss rate. After a terminated contractor's session kept returning 200 for 12 minutes past the HR change, which change brings revocation inside the promised 60 seconds?⌄
MCQ-86A quarterly review flags that the orders service returns any order's details to any authenticated caller: the endpoint takes an order ID and never checks who owns it. The service team's lead replies that the API gateway already enforces tenant isolation, so the finding is a false positive, and asks you to close it. You have a staging account and a second test tenant. What do you do next?⌄
MCQ-89An application role in the analytics account is allowed s3:* on every resource, and the same Terraform module created that role for thirteen teams. The team argues that AWS has encrypted new objects with SSE-S3 and blocked public access on new buckets since 2023, so the provider owns what matters and the wildcard is acceptable. Which option names the control the customer actually owns and states the remediation that holds up?⌄
MCQ-90Your team runs 60 services across 12 teams. A critical advisory lands for a library that 14 of the services ship. Your inventory answers "which services ship version X" in about 20 minutes, but it records nothing about runtime configuration, and the vulnerable code path sits behind a feature flag. Confirming flag state for a similar advisory last quarter took 3 days across 12 teams. A second advisory for the same library is already queued. Which change removes the 3-day step?⌄
MCQ-99A payments team has 80 engineer-hours before the next release and a security debt register of 90 open items. Each item records the operational exposure it creates, a named owner, and a review date. Current state:
| Property | Items |
| --- | --- |
| open | 90 |
| named owner | 22 |
| review date | 14 |
| past review date | 11 |
| appeared as a contributing condition in an incident in the last 12 months | 6 |
All 6 incident-linked items were raised more than 12 months ago, and none of them has a named owner or a review date. Median remediation cost across the register is 10 engineer-hours. A product owner wants the 80 hours spent on a deadline-driven feature instead. Which allocation is most likely to reduce next quarter's incident count?⌄
MCQ-29A detection team proposes a rule that alerts when a host resolves a domain absent from the previous 30 days of corporate resolver logs. The fleet is 8,000 hosts. In the last 24 hours the resolver logs contain 4,800 distinct client addresses, while endpoint telemetry reports all 8,000 hosts online and each connected to at least one destination it had not contacted before. What should the review conclude?⌄
MCQ-28A 900-host estate runs TLS 1.3 on all internal service-to-service traffic. Six of its 40 internal applications pin certificates, and those six also make outbound TLS calls to a partner API. One of the six is the compromised app server that is beaconing to a rare external IP. The database host and the workstation that is DNS-tunnelling are among the 120 appliances or OT devices that cannot run an endpoint agent. Next quarter the team must detect four things: C2 beaconing from that app server to a rare external IP, 4 GB nightly outbound transfers from the database host, DNS tunnelling from the workstation, and exploitation of a known RCE in an unpatched Tomcat instance on a DMZ host that serves plaintext HTTP. A pilot inline TLS-decrypting sensor on the internal segment broke the six pinned applications and was rolled back. Which placement should the team adopt?⌄
MCQ-27A payments platform is segmenting 240 hosts. The team captured 30 days of VPC flow logs, derived a policy from the 46 edges observed, 12 of which appear in no architecture diagram, and enforced it across the card-authorisation segment in one change window. Three weeks later the year-end regulatory reconciliation failed: it runs once a year and reaches the ledger service over a path that exists only during that run. Rollback took 40 minutes. Flow-log retention on the platform is capped at 90 days. Which change to the method would have caught that path before enforcement?⌄
MCQ-25A platform team runs a uniform 30-day patch SLA. Last quarter it was met on 98% of assets, yet an internet-facing VPN appliance was exploited on day 9 of a widely disclosed flaw whose CVSS base score of 7.5 ranked below several unpatched internal findings. The exploited path was not covered by the WAF rules in front of the appliance. Internal libraries have never been exploited inside 30 days, and patching capacity is fixed and fully used. Which change should the review recommend?⌄
MCQ-24A scanner reports 480 open findings across the estate: 60 Critical (CVSS 9.0 or above), 140 High, 280 Medium or Low. Cross-referencing the asset inventory and CISA's Known Exploited Vulnerabilities catalog shows 12 findings on internet-facing hosts, 3 of them in KEV with exploitation observed in the wild: two rated Medium (CVSS 6.5 and 5.8) and one Critical. The other 57 Criticals sit on internal-only hosts with no known exploitation. The team has one sprint of remediation capacity. Which queue should it work first?⌄
MCQ-4An IR team is scoping an intrusion discovered today at a payments company. Hot log retention is 30 days; a 12-month archive exists, but every search against it requires a full restore that takes 48 hours. Forensics dates initial access to 121 days before discovery, persistence to 118 days, lateral movement to 22 days, and exfiltration to 3 days. The incident review is tomorrow morning, 24 hours from now. Which course of action is defensible?⌄
MCQ-75A CSPM scan of 60 AWS accounts returns 4,812 findings. The team has one engineer-week, so the order matters. The scanner ranks by severity; the team ranks by exposure: internet reachability first, then data classification (Restricted > Internal > Public), then privilege (admin > write > read > none). Which remediation order does the exposure ranking produce?
| # | Finding | Scanner severity | Internet reachable | Data class | Privilege |
| --- | --- | --- | --- | --- | --- |
| 1 | Public S3 bucket `cust-exports` | High | Yes | Restricted: 2.1M customer export objects | read |
| 2 | Deploy role assumable from any account | Critical | No | Internal | admin |
| 3 | CloudTrail disabled in 3 production accounts | Medium | No | Restricted | none |
| 4 | Security group 0.0.0.0/0 on port 5432, isolated test host with no route | Critical | No | Public: no data | none |
| 5 | Unused IAM role with write access to production buckets, last used 400 days ago | Low | No | Public | write |⌄
MCQ-73A cloud security team has 120 open scanner findings in one AWS account. Two internet-facing entry points exist: an ALB in front of the app tier, and a public CI runner. The crown-jewel assets are the customer database and the payments service. Among the findings: the ALB forwards unsanitised paths to the app server (medium); the app server's instance role can read the customer database credentials from Secrets Manager (medium); the app server's security group permits the database port (medium); the CI runner has remote code execution (critical) but its network reaches only the build artifact bucket; the other 116 findings sit on hosts with no route to either crown-jewel asset. Given these findings, which prioritisation should the team adopt?⌄
MCQ-65A quarterly scan of a 4,000-host estate returns 340 internet-exposed services: 210 with a named owner and a documented current purpose, 69 with a named owner who confirms the function was retired last quarter, and 61 with no owner in the CMDB that no team will claim. Change policy requires the named owner's sign-off before any service is removed, and edge blocks are temporary mitigations revisited each quarter. The team has one sprint and wants to close as much exposure as it can, permanently where the policy allows. What should it do?⌄