Skip to content
Tutorials Center
My learning

Topics

Lesson 38 of 38 ยท Projects and Practice

Practice: Cloud Security Review and Checklist

Practise a cloud security review: a layered checklist, spot-the-flaw architecture scenarios, a risk scoring exercise and a self-assessment script to plan your next steps.

This final lesson brings the chapter together. You will use a layered checklist to review a cloud workload, practise on small scenarios where you must spot what is wrong, score and rank risks and finish with a self-assessment that turns your answers into a plan. Treat it as a workshop: read each scenario, decide what you would do, then run the code to compare your answer.

All scenarios are invented, and the code runs in your browser. Use the checklist only on systems you own or are authorised to review.

A layered review checklist

Reviews work best when they follow the layers of the system, from identity outward to people and process. For each item ask: is it in place, is it enforced automatically, and is there evidence?

flowchart TB
    A["1. Identity: MFA, least privilege, short-lived credentials"] --> B["2. Network: segmentation, no public admin ports, egress control"]
    B --> C["3. Data: classified, encrypted, backed up, residency known"]
    C --> D["4. Workloads: hardened images, non-root, patched, minimal roles"]
    D --> E["5. Pipeline: scans, signed artifacts, no static secrets"]
    E --> F["6. Detection: logs on and central, alerts with owners"]
    F --> G["7. Response and governance: runbooks, tested backups, compliance evidence"]

Seven layers. Weakness in a lower one can undo strength in an upper one, for example a leaked pipeline key that bypasses perfect network rules.

Layer Questions to ask
Identity Is MFA enforced for every human? Are there wildcard policies? Are old keys and unused accounts removed? Is the root or global admin account locked down and monitored?
Network Are databases and admin ports private? Is traffic between tiers restricted? Is outbound traffic limited and logged?
Data Do we know where sensitive data lives? Is it encrypted at rest and in transit with managed keys? Are backups immutable and restore-tested?
Workloads Are base images small, pinned and scanned? Do containers run as non-root? Do functions have one minimal role each? (containers, serverless)
Pipeline Do builds scan code, dependencies, IaC and images? Are secrets absent from code? Is deployment authenticated with short-lived credentials? (shift left)
Detection Are audit logs on in all regions and stored separately? Are alerts actionable and owned? (logging, SIEM)
Response and governance Is there a tested incident plan? Are compliance controls mapped with fresh evidence? (incident response, compliance)

Exercise 1: spot the flaw

Each description below is a short architecture. Find the problems, then run the code to see one possible list. Try to name at least three issues per scenario before you look.

Scenario A: photo-sharing start-up. Users upload photos through a web app. The app runs on a VM with a public IP. SSH is open to all addresses so that developers can log in from home. The VM uses an access key stored in a .env file to write to a storage bucket. The bucket is public so images can be displayed. Logs are kept on the VM.

Scenario B: internal reporting service. A function reads a database and e-mails a CSV. The function's role is AdministratorAccess because "it was quicker". The database is in a private subnet with TLS required. A dependency was added from a package index last week and not reviewed.

Python Compare your findings with a reference list

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Your list may differ and still be valid. What matters is the reasoning: each problem maps to a layer and has a concrete fix.

Exercise 2: rank the risks

You will rarely fix everything at once. Prioritise using likelihood and impact, both on a scale of 1 to 5. The product gives a score, and you work from the top. Adjust the numbers to match your own judgement, then rerun.

Python Risk register with scoring

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Ties are broken by effort, so cheap fixes with the same score come first. Quick wins build momentum and reduce risk early.

Exercise 3: a self-assessment

This script scores your own organisation from 0 (nothing) to 3 (automated and monitored) for each layer, then prints your weakest areas and the lesson to revisit. Change the numbers in answers to match reality. Be honest: the value is in the gaps.

Python Self-assessment and improvement plan

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Use the plan as the start of a 30-60-90-day roadmap: in 30 days fix the quick wins, in 60 add automation, in 90 test your response and your detections.

Habits that last

  • Automate the checks so reviews are not a yearly event: policy as code, scanners in pipelines, posture monitoring.
  • Assume breach. Design so that one stolen credential or one compromised workload cannot reach everything: least privilege, segmentation, short-lived credentials, logs elsewhere.
  • Practise. Tabletop exercises, restore tests and authorised simulations show gaps while the stakes are low.
  • Keep learning. Provider documentation, the CIS Benchmarks, OWASP, the MITRE ATT&CK matrix and your provider's well-architected security guidance are good next steps.
  • Stay within the law and your authorisation. Test only what you own or are explicitly permitted to test, and report vulnerabilities in other people's systems through their disclosure process.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Why review a workload layer by layer, from identity to governance?

    Attackers use the weakest layer. A systematic review prevents blind spots.
  2. 2. Which is the best first fix in the risk-ranking exercise?

    High likelihood and impact together with low effort give the biggest risk reduction early on.
  3. 3. What does 'assume breach' encourage you to do?

    If a breach is likely to happen eventually, you design so that it stays small and detectable.
  4. 4. Where should you run security tests and scans?

    Unauthorised testing is illegal in many places and harmful. Use your own accounts and test environments, and follow responsible disclosure for anything else.

Summary

  • Review cloud workloads layer by layer: identity, network, data, workloads, pipeline, detection and response with governance.
  • Practise spotting flaws in realistic descriptions, and tie every problem to a layer and a fix.
  • Rank risks by likelihood and impact, and take cheap high-value fixes first.
  • Use a self-assessment to build a 30-60-90-day plan, automate checks and rehearse your response.
  • Stay inside your authorisation at all times.

Next: Back to the Cloud Security track