Cloud misconfiguration review: how to find risky settings
How to review cloud misconfiguration risks in Terraform or CloudFormation: public storage, open admin ports, public databases and secrets in code.
Free app. Runs on your Mac with the Claude or Codex plan you already have.
What a cloud misconfiguration review is
A cloud misconfiguration is a setting that leaves a resource more exposed than intended: a storage bucket readable by anyone, an administrative port open to the internet, a database with a public address, or a password written into code. A cloud misconfiguration review reads those settings, usually in the infrastructure as code (IaC) files that create them, such as Terraform or AWS CloudFormation, and lists each risky one with a change an engineer can review.
The method below is how an experienced cloud security engineer reviews AWS declarations by hand. It is deliberately bounded: a review of files shows what the code declares, not what is running, so every finding is a draft for the infrastructure owner to verify against the live account before anything changes. Each step says what to do and why, then how 10folds carries it out.
What it does for the business
- Fix the riskiest settings firstA short queue that puts production policy violations and possible secrets first sends engineering time where exposure would hurt most. Lower-risk questions wait without being lost.
- Catch problems before they deployReviewing declarations finds risky settings at the point where they are cheapest to change: in a pull request, before anyone depends on them.
- Prepare for an audit or a customer reviewA dated record of what was checked, what was found and what remains unknown gives auditors and customers a clear picture of your process without overstating it.
- Changes engineers can trustEach proposed change comes with its prerequisites and how to roll it back, so tightening access does not break a service that depended on it.
Cloud misconfiguration, step by step
Map the declarations and the policy
A setting you cannot resolve is unknown, and unknown is not safe.
What to do
- List the files and resources for one environment and revision, separating production from examples and tests.
- Record how variables, conditions, count and for_each resolve, and where module contents are missing.
- Collect the company's approved exposure rules, the services meant to be public and any dated exceptions.
- Treat an expired exception as something to review, not as approval.
- Mark resource types and other clouds you will not review as out of scope.
Why
Most false alarms in Terraform security reviews come from reading an example file as production, or a variable default as the deployed value. A missing setting at bucket level does not prove there is no protection at account level. Knowing which services are meant to be public stops the review from recommending changes that would break them.
Check public storage settings
One false protection breaks a policy. It does not yet prove data is exposed.
What to do
- Read the four S3 public access block settings for each bucket separately.
- Read bucket policies and ACLs (access control lists) that grant access.
- Compare against the company rule, for example that all four protections must be on.
- Ask the owner to confirm any bucket meant to serve public content.
Why
Public storage is among the most common cloud misconfigurations and among the most damaging. The four settings do different things, so one switched off is a different risk from all four. A bucket hosting a public website needs a recorded decision, not a blanket lock that breaks the site. Account-level settings can also block public access across every bucket, which is why each layer is read separately.
Check admin ports and public databases
A restrictive rule does not cancel a permissive one beside it.
What to do
- Read every security group ingress rule, for both IPv4 and IPv6 ranges.
- Flag SSH (port 22), RDP (port 3389) or all-protocol rules open to the whole internet.
- Check complete port ranges: a range from 0 to 65535 includes port 22.
- For each database, compare the public accessibility setting with the intended architecture.
Why
Security groups add rules together, so an open IPv6 rule stays open even when the IPv4 rule is narrow. Web ports open to the world are normal for a web server and are not administrative exposure. A database marked publicly accessible is a concern in its own right, but reachability also depends on routes, subnets and authentication, which files alone may not show.
Look for secrets in code
Report where a secret is. Never repeat what it is.
What to do
- Search for passwords, keys and tokens written directly into variables, defaults or resources.
- Tell real-looking credentials apart from placeholders and examples.
- Remember that a sensitive marker hides a value from output, not from state files.
- Send suspected real credentials to the owner for verification and rotation.
Why
Secrets in code end up in version control history and in state, where many people can read them. Copying a secret into a findings report spreads it further. Testing a found credential to see if it works is itself an access attempt, so it is left to the owner. Placeholder values in example files look like secrets but are not, and flagging them wastes the owner's time and trust in the review.
Propose changes, review them and hand over
When a critical value is missing, write the question, not a placeholder patch.
What to do
- Write each change as before and after, with the owner, prerequisite and rollback.
- Ask the owner for values you do not know, such as the office network range.
- Have a second person reopen every cited line and check IPv6, unresolved values and redaction.
- Keep a coverage table: what was examined, what is unknown and what was out of scope.
- Deploy only after the owner confirms the approved access paths still work.
Why
An invented network range in a patch can lock a team out of its own servers. A coverage table shows the limits of the review, so a short list of findings is not read as a clean bill of health. Least privilege is the goal, but each step toward it must keep the access the business actually needs.
Common mistakes
- Treating unknown as safeAn unresolved variable may hold the risky value in production.
- Forgetting IPv6A rule open to ::/0 is open to the internet even when IPv4 is narrow.
- Pasting secrets into the reportThe report becomes one more place the credential leaks from.
- Locking every bucketSome content is meant to be public. Ask before breaking a site.
- Calling a file review proof of securityIt shows what code declares, not what is deployed.
What a good result looks like
- A stated scope: environment, revision and the files and resources examined.
- Each finding linked to the exact file and line.
- A disposition for every check, including Unknown and Unreviewed.
- Proposed changes with owner, prerequisite and rollback.
- No secret values anywhere in the output.
And then what? Using the result
- Verify each finding against the live account before changing it.
- Rotate any credential flagged as possibly real.
- Apply approved changes through your normal pull request process.
- Answer the owner questions so Verify items can be closed.
- Add IaC scanning to your pipeline to catch the same settings earlier.
When to run it
- Every monthA regular review of production declarations catches settings that drifted in through small changes.
- Before an audit or a customer security reviewA dated, scoped record of what was checked helps answer questions without overstating.
- Before a large infrastructure change shipsReviewing the new declarations finds exposure while it is still a pull request.
- After a credential leak scareThe secrets check shows where else credentials may be written into code.
Questions people ask
What is a cloud misconfiguration?
A cloud misconfiguration is a setting that exposes a resource more than intended, such as public storage, an admin port open to the internet, a publicly accessible database or a secret written into code.
How do you review Terraform for security?
Fix the environment and revision, resolve variables and modules, check each resource against clear rules for storage, network ingress, database access and secrets, record unknowns, and propose reviewed changes with rollback.
What are the most common cloud misconfigurations?
Public storage buckets, security groups open to the internet on administrative ports, databases with public access and credentials stored in code or state are among the most frequently cited.
Does a scanner replace a manual review?
No. Scanners check many rules quickly and suit a pipeline. A manual review adds context: which services are meant to be public, which exceptions are approved and which values are unknown. Many teams use both, often alongside the CIS Benchmarks.
Does 10folds connect to my AWS account?
The review method reads the Terraform or CloudFormation files you supply. It does not discover live resources, probe them, use credentials or write to any cloud API. Every finding is a draft for the infrastructure owner to verify against the live deployment.
Which clouds and resources does it cover?
AWS public storage settings, security group admin ingress, database public access and secrets in code. Other clouds and resource types are listed as unreviewed rather than judged by AWS rules.
More jobs in Developers
Review cloud settings for risks, on your own Mac.Free app. Runs on your Mac with the Claude or Codex plan you already have. Your files stay on your computer.
Download for Mac