AWS Security Review: Find the Attack Paths, Not Just a List of Alerts

A long list of findings does not tell you what to fix first. How attack paths change the order, what a good AWS security review covers, what read-only should mean, and a plan for week one.

AWS Security Review: Find the Attack Paths, Not Just a List of Alerts

Most AWS accounts that have never had a security review do not have one problem. They have a long list of findings, and no way of telling which five matter.

The short answer: an AWS security review is only useful if it tells you what to fix first. Sorting findings by severity does not, because severity ignores whether a weakness connects to anything an attacker wants. Linking findings into attack paths, from the internet or an outside account to administrator access or sensitive data, changes the order: the fix that breaks the most paths goes first, and many "high" findings that sit on no path can wait. Run it from a role that is truly read-only, treat what it could not see as unknown, not clean, and re-check after you fix.

For CTOs, heads of engineering and founders running on AWS without a dedicated security team.

Why a list of findings does not tell you what to fix first

A typical AWS security assessment produces a list: an open security group, an access key nobody has used in months, an unencrypted volume, a role with more permissions than it needs. The instinct is to sort by severity and work down. That order is usually wrong, for three reasons.

  • Severity is judged one finding at a time. An unencrypted disk attached to nothing reachable is rated the same as one behind an exposed server, though only one of them is on anyone's route in.
  • Dangerous combinations are made of medium findings. A function anyone can call is not alarming; nor is a broad role. A public function running as that role is a way into the account, and neither finding says so.
  • A small team cannot fix everything at once. It can fix five this week. The question the review has to answer is which five.

How attack paths change the order

An attack path joins findings into a chain from an entry point to a target. Entry points are the ways in: the internet, a principal in another AWS account, or a credential that has leaked. Targets are what an attacker wants: administrator-equivalent access to the account, or a store of sensitive data.

  1. Severity becomes contextual. A finding that sits on a path matters more than its rating on its own; a high finding on no path matters less.
  2. Fixes are ranked by the paths they break. Several paths often run through the same step, typically an over-broad role. Narrowing that one role can break more paths than ten fixes elsewhere, so it goes to the top.
  3. Uncertainty is visible. Some steps depend on what configuration cannot confirm, such as a flaw in a function's code. A good review marks those paths possible, not proven, and names the assumption.

That is what AWS attack path analysis is for: it does not replace the list, it gives it an order.

How to read an attack path: an example

The example below is illustrative. It is not taken from any real account.

StepWhat the review sawWhy it is a step
Entry: the internetSSH open to anywhere on a web server with a public IP addressAnyone can attempt to reach the server
Step 2: the serverThe server still allows the older instance metadata service (IMDSv1)A request the server can be tricked into making can read its role's temporary credentials more easily
Step 3: the roleThe server's instance role has administrator permissionsWhoever holds those credentials holds the account
TargetAdministrator-equivalent accessThe path is complete

How to read it:

  • Status: possible. Step 2 assumes the attacker can make the server send requests. Configuration cannot prove that, so the path is a risk to remove, not evidence of a breach.
  • Three fixes, each enough on its own. Closing SSH to the internet, requiring the newer metadata service, or narrowing the role would each break this path. If the same role also sits on two other paths, narrowing it breaks three, and it goes first.
  • What drops down the list. An unencrypted volume on no path keeps its finding but loses its urgency: worth fixing, not this week.

AWS security review checklist, by area

A useful AWS security audit checklist covers six areas, each of which tends to play a particular part on a path.

AreaWhat to checkHow it shows up on a path
NetworkAdmin ports open to the internet in security groups or network ACLs; all-traffic rules; rules on default security groups; public load balancers and outdated TLS; Lambda function URLs and API Gateway routes without authentication; DNS records pointing at deleted resourcesUsually the entry point
IdentityRoot access keys and root MFA; users without MFA; old keys, and unused users with active keys; administrator principals; roles that can escalate their own privileges; cross-account trust without an external ID; admin instance roles; external access found by IAM Access AnalyzerUsually the step that turns a foothold into control
DataPublic S3 buckets and account-level Block Public Access; public snapshots and AMIs; queue, topic, key and registry policies open to anyone; unencrypted EBS and RDS; open OpenSearch domains; CloudFront origins reachable directlyUsually the target, sometimes the entry
ServersInstance metadata v1 still allowed; public IP addresses; instances not managed by Systems Manager; deprecated Lambda runtimes; public EKS endpoints; publicly accessible RDS; outdated AMIs, database engines and operating systemsThe middle of a path: where an entry becomes access to a role
LoggingCloudTrail across all regions with log file validation; GuardDuty on; VPC flow logs; AWS Config recording; activity in regions you do not use; KMS key rotationRarely a step; decides whether you would notice anyone walking one
Credentials and stored secretsStored secrets that are never rotated; databases still on a static master password; an inventory of which credentials exist at allA leaked or long-lived credential is an entry point in its own right

Two optional areas go deeper and need more access: known vulnerabilities on servers and packages, and secrets left in configuration, such as keys in function settings, server user data or deployment templates. AWS's own IAM security best practices are a sound reference for the identity rows; they recommend MFA, least privilege and regularly removing unused users, roles and credentials.

What read-only should mean, and why it matters

A review that adds a new way into your account has defeated its own purpose, so check "read-only" line by line.

  • Read-only in IAM. Policies that cannot create, change or delete anything, such as AWS's managed SecurityAudit and ViewOnlyAccess.
  • Explicit denies on what you would not hand over. Read-only still allows a lot, so add denies on secret values, S3 object contents, parameter values and decryption.
  • Only the reviewer can use it. Trust only the reviewer's named identities, require an external ID tied to your connection, and keep sessions short.
  • Read-only in the tool's code, too. A tool that refuses any non-read call before it leaves the process does not depend on the IAM policy being perfect.
  • Revocable in one step. If the role came from a CloudFormation stack, deleting the stack ends access at once.
  • Honest about blind spots. Anything the review could not read is a gap, with the permission it needed, never a pass.

What usually goes wrong in an AWS security review

  • Fixing by severity count. The team closes thirty medium findings and leaves the one role that three paths run through.
  • Treating "not checked" as "clean". A region the scanner could not read looks like a clean result unless the report says otherwise.
  • Fixing without re-checking. A finding is only resolved when the check that raised it has run in full again.
  • A one-off that is never repeated. A review is a snapshot, and its value decays with every deploy.

A cybersecurity risk assessment is the larger exercise across vendors, people and compliance; an AWS security review is one input to it.

How this compares with AWS Security Hub

AWS offers two related services. The comparison below uses AWS's own documentation.

AWS Security Hub CSPMAWS Security Hub (latest)A point-in-time attack path review
How it runsInside your accounts, continuous checksInside your accounts, enabled per account or across an organisationWhen you choose, from a read-only role
What it checks againstStandards: AWS Foundational Security Best Practices, CIS, PCI DSS, NISTCorrelates CSPM, Amazon Inspector, Macie and GuardDuty signals into exposure findingsMisconfigurations linked into paths to admin access or data
Attack pathsNot described in the CSPM overviewYes, an attack path graphYes, with fixes ranked by paths broken
PrerequisitesAWS Config enabled and recording, for most controlsUses findings from the services it correlatesA role in the account
ResponseAutomation rules, EventBridgeAutomated response workflows and ticketing integrationsA report for a person to act on
Cost30-day free trial, then billed by AWSSee AWS's Security Hub pricingDepends on the provider

Sources: Security Hub CSPM and Security Hub documentation.

If you need continuous monitoring, compliance scores or automatic response across an organisation, Security Hub is built for that. If you need an ordered answer to "what do we fix first in this account?" without switching services on, a point-in-time path review is the shorter route. They are not exclusive.

Your first week

  1. Day one: list what you have. Every AWS account, its owner and the regions in use.
  2. Day one: connect read-only and scan. One account at a time, every enabled region.
  3. Day two: read the paths before the list. Note each path's entry point, target and whether it is proven or possible.
  4. Days two to four: fix the top of the ranking. Usually: narrow a role or two, close an internet-facing door or two, deactivate unused keys, and fix root MFA and root keys if open.
  5. Day four: deal with the gaps. Grant the read each one needed, or record it as unknown.
  6. Day five: scan again. Confirm what is resolved, what is new and what could not be re-checked. Then decide whether you also need continuous monitoring.

Where AWS Cloud Security fits

AWS Cloud Security is this review as a product on Tomorrow Central. You connect an account with a read-only role from a CloudFormation template; only Tomorrow Central can use it, with a code tied to your connection, for an hour per session at most. A scan runs 53 read-only checks across the six areas, in every enabled region, links what it finds into attack paths and ranks the fixes by how many paths each one breaks. Severity is adjusted for the paths a finding is on, and anything the scan could not read is listed as a gap. Two optional scans, for known vulnerabilities and for secrets left in configuration, are off until you allow them on the role; a secret scan keeps only a fingerprint and a masked preview of what it finds, never the value.

Its limits: it scans one account at a time, when you run it; it does not evaluate service control policies or permission boundaries, so permission steps are marked possible; it does not score compliance, monitor continuously or fix anything for you; and it is not a complete security assessment.

Questions people ask

What is an AWS security review?

A structured read of one AWS account's configuration to find what an attacker could use: who can reach what from the internet, which identities and keys are too powerful, which data is exposed and what is not being logged. A good review ends with an ordered list of fixes, not just a list of findings.

What should an AWS security audit checklist cover?

At least six areas: network exposure, identity and access keys, data stores and their policies, servers and runtimes, logging and detection, and stored credentials and secrets. The checklist matters less than the order you fix things in, which is where attack paths help.

How is an attack path review different from AWS Security Hub?

Security Hub runs inside your accounts, checks them continuously against standards such as AWS Foundational Security Best Practices and CIS, relies on AWS Config for most checks, and in its latest version correlates findings into exposure findings with an attack path graph. A point-in-time path review runs when you choose from a read-only role and ranks fixes by the paths they break. They answer different questions, and many teams will want both.

How often should we run an AWS security review?

Before an audit, a launch or a customer's security questionnaire, after a significant change to the account, and after each round of fixes to confirm they held. If you need to know within hours when something changes, you need continuous monitoring as well.

Does an AWS security review replace a penetration test?

No. A configuration review reads how the account is set up; it does not test, probe or attack anything. A penetration test tries to exploit what is running. The review tells you where a tester's time is best spent.

Key takeaways

  • A list sorted by severity is not a fix order; linking findings into paths from an entry point to admin access or data is.
  • Rank fixes by the number of paths they break. One over-broad role is often worth more than ten isolated findings.
  • Read-only should be enforced in IAM and in the tool's code, revocable in one step, and honest about what it could not see.
  • Security Hub and a point-in-time path review answer different questions; choose by whether you need continuous monitoring or an ordered fix list now.

Where to start

To run this review on one of your accounts, request access to AWS Cloud Security on Tomorrow Central; it is switched on for each organisation while it is new. If you would rather have the review run, interpreted and the fixes carried out for you, see ITMTB's cybersecurity services or contact us.

Sources

  1. AWS, Introduction to AWS Security Hub CSPM, AWS Security Hub User Guide (accessed 16 November 2026)
  2. AWS, Introduction to AWS Security Hub, AWS Security Hub User Guide (accessed 16 November 2026)
  3. AWS, Security best practices in IAM, AWS Identity and Access Management User Guide (accessed 16 November 2026)
  4. Tomorrow Central, AWS Cloud Security

Get a Structured Cybersecurity Risk Assessment

We work with mid-to-large enterprises to map cloud exposure, identity risk, vendor dependencies, infrastructure vulnerabilities, and compliance readiness — and deliver a remediation roadmap with named owners, not just a list of findings.

Explore More Insights

Cybersecurity Risk Assessment for Enterprises: Framework, Operational Risks, and Implementation Guide

Cybersecurity Risk Assessment for Enterprises: Framework, Operational Risks, and Implementation Guide

Read More
AWS Cost Optimization: How to Find Cloud Resources You Are Paying For but No Longer Need

AWS Cost Optimization: How to Find Cloud Resources You Are Paying For but No Longer Need

Read More
AI Agents and Enterprise Cybersecurity: Risks, Governance, and Operational Controls

AI Agents and Enterprise Cybersecurity: Risks, Governance, and Operational Controls

Read More