Skip to content
Tech Interview Prep
Technical interview guide

Incident Response Basics

The standard phases of handling a security incident, from detection through lessons learned.

Read
27 min
Practice MCQs
25
Interview Q&A
25
Edition
v3
Editorial status
Reviewed

Scope: NIST SP 800-61 Rev. 3 (2025), CSF 2.0-aligned guidance, and official CISA playbooks accessed 2026-08-31..

Curated: · Written: · Reviewed:

25 questions

MCQ-2A security team's incident response plan is organized as four sequential phases: Preparation; Detection and Analysis; Containment, Eradication, and Recovery; Post-Incident Activity. Over six months, the same misconfigured S3 bucket was the initial access vector in three separate incidents. Each time, the Post-Incident Activity phase logged lessons-learned actions, but no one owned them and none were tracked to completion before the next incident. The team is revising the plan to align with NIST SP 800-61 Revision 3. Which change to the plan structure addresses the recurring failure?
MCQ-1A SIEM rule fires on host `web-12` (internet-facing, not in the cardholder-data environment): outbound transfer of 140 MB to 198.51.100.9 between 14:02 and 14:07, against a rule threshold of 100 MB in 5 minutes. Triage shows: `web-12`'s crontab has a 14:00 daily backup that transfers 130–160 MB to 198.51.100.9; the destination's TLS certificate matches the backup vendor; the session used the `backup` service account, the only account that runs the job; no failed logins, new processes, or file-integrity changes. The org's criteria define an incident as an event that actually or potentially jeopardizes confidentiality, integrity, or availability, or violates policy. What is the correct triage outcome?
MCQ-5At 14:02 you declare SEV-2 for a CI runner after its GitHub App token is used from IP 203.0.113.7 to clone the private repo `acme/docs`. At 14:20 you check the IP: it is your own `us-east-1` egress NAT gateway, and the clone matches the runner's scheduled `docs-sync` job that ran at 13:50. The token has `contents:read` on `acme/docs` only, the runner holds no cloud credentials, and no other use of the token appears in the audit log. Your rubric: SEV-1 = a credential that reaches production data, payment systems, or IAM; SEV-2 = non-production credential compromise with confirmed external use; SEV-3 = suspicious activity with no confirmed credential use. What is the correct severity and next action?
MCQ-16During a postmortem, two log lines are merged into one timeline: host A (clock 47 s fast) logged 'token issued 10:00:47' and host B (clock 12 s slow) logged 'token used 10:00:05'. Read raw, the use appears 42 s before the issue, and a responder reads that inversion as a replayed token. Which reading belongs in the timeline?
MCQ-17At 14:02 a checkout service's p99 latency jumps from 40 ms to 180 ms. The on-call engineer sees a Grafana alert: 'DB connection pool active = 20/20' and a deploy at 13:58. CPU on the DB is 88% idle, and pg_stat_activity shows 18 idle-in-transaction sessions. The incident channel already has two responders. The engineer forms the hypothesis: 'The 13:58 deploy introduced a connection leak that exhausted the pool.' What should the engineer record in the incident channel as the working hypothesis?
MCQ-18At 02:14, EDR flags FIN-WS-14 making outbound TLS to 203.0.113.77:443. The IP is in a commercial threat feed with confidence 60, first seen 2019, last seen 2021, tagged "Cobalt Strike C2". The IP resolves to a shared hosting provider that hosts 4,000 domains. EDR shows chrome.exe (signed, parent explorer.exe) initiated the connection, with no child processes and no persistence. The SOC has 15 minutes before shift change. What should the analyst do?
MCQ-13An API host compromised through a vulnerable dependency is rebuilt from a clean image at 06:00, its credentials rotated, and it rejoins the pool. The team runs 72 hours of heightened monitoring. At 06:00 on day 2, the host's outbound DNS query rate rises from 4/min to 60/min. 55 of the 60 queries are for distinct randomly generated domains, and all 55 return NXDOMAIN. The host's CPU, memory, request rate, and latency are unchanged. What does this pattern most likely indicate?
MCQ-4A SEV1 checkout outage has just started. The on-call database engineer is the first responder in the channel and is the only person with credentials to run the failover, which takes about 15 minutes. Two other engineers join the channel. What should the database engineer do?
MCQ-19During an intrusion, responders confirm the attacker holds an active session in a responder's mailbox and can read and alter the corporate chat tenant. Which coordination instruction should the incident commander give?
MCQ-20At 09:00 Monday an engineer finds that an unauthenticated API endpoint has been returning customer records (name, email, hashed password) to anyone who guessed a numeric ID. Logs show the endpoint was reachable from the internet from Thursday 18:00; it is being disabled now and will be closed within 20 minutes. About 40,000 of the affected customers are in the EU; the company is incorporated in Delaware and has no EU establishment. What should the incident commander do about legal and privacy?
MCQ-21An analyst is drafting the 72-hour regulatory breach notification for a ransomware incident. Containment is complete. Three attribution indicators are on the whiteboard: (1) a C2 domain registered 11 days before the intrusion, (2) a binary whose embedded PDB path contains the string 'apt29_tools', and (3) a ransom note whose phrasing matches a public report describing Group X. The team's rubric scores one independent source as low confidence, two as moderate, and three as high. The PDB string and the ransom-note phrasing both come from that same public report. What should the notification say about the actor?
MCQ-7An analyst images a compromised server's disk at 14:00, records the image hash, and logs the drive into a shared evidence locker. The log shows a second analyst checked the drive out at 16:00; the return line at 22:00 is unsigned, and that analyst has since left the company, so the six hours the drive spent outside the locker are unaccounted for. The image hash still matches the value recorded at 14:00. Opposing counsel challenges the image's integrity. What is the most accurate assessment?
MCQ-6A compromised Linux server has 64 GB of RAM. The attacker's implant is fileless: it exists only in memory, and the disk holds no implant artifacts. The server's 1 TB data volume is LUKS-encrypted and the volume key lives only in kernel memory. The response team has already cut the server's network link, so exfiltration has stopped. The implant reboots the host 10 minutes after it loses its C2 heartbeat, and the link has been down for about 30 seconds. The team's memory acquisition tool reads RAM at 160 MB/s. What should the responder do first?
MCQ-9A ransomware crew beacons from 40 hosts to a single C2 domain. The team blocks that domain at the egress proxy; beaconing stops. Three days later, with the block still in place, 9 of the 40 hosts beacon to a new domain through the same proxy. No host has been reimaged and no credential has been rotated. Which conclusion does this evidence support?
MCQ-10An API issues 15-minute RS256 access tokens that are validated by signature alone, plus opaque refresh tokens held in Redis and rotated on every use (the previous token is deleted at redemption). At 14:02 a user's refresh token is confirmed copied from a compromised laptop; at 14:03 the attacker redeems it and receives a new access token and a new refresh token. Which containment action stops the attacker from obtaining further access tokens?
MCQ-8A ransomware detonation has encrypted 40 of 900 endpoints and is still encrypting two file servers. The attacker holds a live domain admin account, EDR telemetry is still streaming from the affected hosts, and offline backups are six hours old. Which containment step fits this situation?
MCQ-15At 02:10, EDR shows a radiology workstation running a ransomware binary with open SMB sessions (TCP 445) to three file servers. The domain controller is not encrypted, but the same workstation holds a live domain-admin session. The Veeam repository is domain-joined on the same /24 as the file servers. Medication cabinets authenticate to AD every 15 minutes and lock if they cannot reach a domain controller for 30 minutes. Which first action best contains this incident?
MCQ-3A payments team must show a 4-hour RTO for its 4 TB primary database after a ransomware incident. Its preparation evidence: an IR plan approved 14 months ago; a ransomware tabletop 9 months ago scored 'pass'; nightly full backups with 100% job success for 12 months; a DR runbook last executed 2 years ago. The only measured restore, from that 2-year-old drill, restored 4 TB at a sustained 160 MB/s. Which conclusion does this evidence support?
MCQ-14A payments service runs PostgreSQL 16 with a weekly full base backup (Sunday 01:00, 900 GB) and continuous WAL archiving to S3 (`archive_timeout = 60`, WAL uncompressed, ~12 GB/hour). The DR runbook claims a 1-hour RTO. In a drill, restoring the base backup from S3 runs at 600 MB/s and takes 25 minutes; WAL replay runs at 90 MB/s. The rehearsed incident is a corruption found Wednesday 12:00, restored to Wednesday 11:59. 83 hours after the Sunday base backup. What restore time should the team plan for, and what dominates it?
MCQ-11Triage of a production database host confirms the attacker held root for about six hours and cleared auditd and shell history, so the team cannot enumerate what persistence was installed. A signed golden image, a tested restore procedure, and a verified backup taken 20 minutes before the earliest confirmed attacker activity are all available. What should the team do?
MCQ-12On 14 March an EDR alert shows a web shell on app-07. Log review puts the attacker's first successful login at 02:11 on 2 March, using a deploy key stored in CI; from app-07 the attacker could also read the Postgres password in the host config. Nightly full backups run at 01:00, and every order write is also appended to a Kafka topic retained for 30 days. Which recovery plan returns app-07 to a known-good state?
MCQ-22A mistyped feature-flag value pushed to all 240 production nodes at once took checkout down for 38 minutes. The deploy tool applies config atomically: no canary stage, no schema validation. The draft postmortem names 'operator typo' as the root cause and its only action item is that the operator attend config training. Two earlier incidents this quarter came from the same class of typo and were both closed with 'be more careful' actions. Which revision to the draft makes the review blameless but accountable?
MCQ-23An incident review's only action item was "raise the database connection pool from 20 to 50 per instance". It was closed when the config change merged. The service runs six instances, and the database's max_connections is 200. Two weeks later, under load, the service fails again: this time Postgres returns "FATAL: sorry, too many clients already". What does the recurrence show about the action item?
MCQ-24Six weeks after an on-call bonus was gated on a single incident metric, a team's reported incident count is down 35% while customer-visible error rates are flat, and 12% of closed incidents were reopened within a week. Which metric was most likely the gate?
MCQ-25At 02:14 a config deploy set the checkout service's connection pool max to 4 (from 40). p99 latency went from 90 ms to 4.2 s and the 5xx rate from 0.1% to 18%. A rollback at 02:31 restored both to baseline by 02:40. During the 17-minute window 3,400 checkouts failed with 503, and 210 of those had already captured payment before failing. The bad config v412 is still queued in the release pipeline for the 09:00 deploy. At 02:45 the incident commander asks whether the incident can be closed. Which decision is correct?