Who this is for
This Cloud security review checklist for small teams is for companies that run one to five cloud accounts and do not have a full-time security engineer. The usual reader is a founder, CTO, IT lead, or the one engineer who also owns backups. You can run the review without buying a tool, but you need read-only access and two hours of quiet time.
The checklist fits two moments. One is the quarterly housekeeping pass, before the account drifts too far from what you intended. The other is the week before a customer sends a security questionnaire or an auditor asks for evidence. It is not a penetration test and it will not find every application bug. It finds the account-level problems that cause most avoidable incidents: stale identities, open ports, missing logs, and backups nobody tested.
If your team has never done this, start with one account. A small review that finishes beats a broad review that stalls.
Before you begin
Write down the scope before you touch a console. List every cloud account, subscription, project, and tenant, including the test account someone created for a demo and forgot. Add the regions in use, the billing owner, and the person who can approve changes. Put the list in a shared document with the review date.
Get read-only access for the reviewer. On AWS, the SecurityAudit managed policy plus ViewOnlyAccess covers most checks. On Azure, use Reader and Security Reader. On Google Cloud, use a custom role with viewer permissions. Do not use a root or global admin account for the review. If you must use a break-glass account, rotate its credentials after the review and record who used it.
Set a timebox and a definition of done. Ninety minutes is enough for a first pass on one account. Each checklist item gets one of three results: pass, fail, or not applicable, plus a short note and a link to evidence. Evidence can be a screenshot, an exported policy, or a command output saved to the shared folder. Without evidence, the next review starts from zero.
The checklist
Work through the items in order. The early items cover identity, which is where most cloud incidents start. The later items cover exposure, logging, and recovery. If an item does not apply, write why, so the next reviewer does not re-open the same question.
Keep the review read-only. If you find a fix that takes less than five minutes, such as deleting an unused access key, make the change in a separate ticket with a second person watching. Mixing review and repair makes it hard to tell what changed and why.
- Inventory every cloud account, subscription, project, and tenant, and record a business owner and billing contact for each.
- Confirm that root or global admin accounts use hardware MFA, have no access keys, and send sign-in alerts to a monitored inbox.
- Review human users in your identity provider, disable departed staff, remove stale guest accounts, and require MFA for every active user.
- List every role with admin or owner permissions, confirm the need, and remove permanent admin rights from daily working accounts.
- Find access keys and service account keys older than 90 days, rotate or delete them, and move workloads to short-lived credentials where the platform supports it.
- Check public storage buckets, databases, and load balancers, and confirm that each public endpoint is intentional and owned.
- Review security group and firewall rules for management ports open to 0.0.0.0/0, such as SSH, RDP, and database ports.
- Confirm encryption at rest for storage, databases, and backups, and record who can read or manage the keys.
- Verify audit logs are enabled and shipped to a central destination that a compromised workload account cannot delete.
- Set alerts for root sign-in, IAM policy changes, billing spikes, and repeated failed console logins, and route them to a named person.
- Test a restore of one critical database, record the time and steps, and fix anything that makes the restore unclear.
- Review third-party OAuth grants, marketplace apps, and integration keys, and remove access that no current process uses.
- Check that each account has a budget alert and an owner, because unused resources often hide old credentials and forgotten test data.
What good looks like
A good review produces a short list, not a hundred-page report. Every fail has one owner, one due date, and one concrete fix, such as removing the 0.0.0.0/0 rule on port 22 in the staging security group. The fix should be small enough to finish in a normal week.
Evidence lives in one place and is named by date and account. A new engineer can open the folder and see what passed, what failed, and what was accepted as a known risk. Accepted risks have a reason, an approver, and an expiry date, so they do not become permanent.
The second review takes less time than the first. You spend less time finding accounts and more time checking changes. If the same finding shows up three reviews in a row, the problem is ownership, not tooling.
What to do next
Fix the highest risk items first. Public exposure and admin identity problems come before cosmetic tagging cleanup. Put the rest into the normal backlog with dates, not a separate security project that nobody runs.
Schedule the next review now, before the calendar fills. A quarterly cadence works for most small teams, with an extra pass after a major migration, a funding round, or a new customer contract that includes security terms.
If no one on the team can own the review, or if a customer wants an outside view, an option is the Cloud security review from Cloudverg at $6,500/month. It is a fixed scope read-only review of up to five cloud accounts over two to three weeks, with identity, exposure, encryption, logging, and backup checks, findings ranked by risk, and a 60-minute readout. Compare that with hiring a contractor for a few days or asking your cloud provider's support team for guidance, and pick what fits your budget and timeline.
Frequently asked questions
How often should a small team run a cloud security review?
Quarterly is a practical default, with an extra review after a major architecture change or a new customer security requirement. The review should take less time as your evidence and owners mature.
What should we fix first after a cloud security review?
Start with public exposure and admin identity. Close open management ports, remove public storage, disable unused admin roles, and turn on root sign-in alerts before you touch lower risk findings.
Can we run this checklist without a dedicated security engineer?
Yes, if one person owns the review and has read-only access. Use the checklist to keep scope fixed, save evidence, and bring in outside help only when a finding needs skills the team does not have.