# Attack path analysis finds the routes to your assets; removing them is a separate job

Guide, 10 Nov 2026, read 10 Nov 2026

**The short answer.** Attack path analysis builds a graph of cloud assets, identities and network links, then searches it for routes from an entry point to a valuable resource. It ranks risks by reachability. The output is a finding, and a finding becomes a fix only when someone changes the policy or the network.

## What attack path analysis is

Microsoft's documentation defines an attack path as a series of steps a potential attacker uses to breach an environment and reach its assets. It starts at an external entry point, such as an internet-exposed vulnerable resource, and follows available lateral movement, such as attached identities with permissions to other resources. Microsoft adds that attack path analysis covers storage accounts, containers, serverless environments, unprotected repositories, unmanaged APIs and AI agents.

To find these routes, Defender for Cloud builds what Microsoft calls a cloud security graph from asset inventory, connections, lateral movement possibilities, internet exposure, permissions, network connections and vulnerabilities. It then searches the graph for exploitable entry points and the steps from there to critical assets. Other platforms, including Wiz and Tenable, describe comparable graph or path models in their own documentation. Each platform builds its own graph, so the same estate can produce different paths in different tools.

## What it finds that a permission list does not

A permission list answers what one identity can do. A path answers how an attacker gets to that identity and what the identity opens next. Microsoft's CIEM page gives an example: an attacker compromises a service principal with excessive rights and moves laterally from a compromised resource to a sensitive database. Each link is ordinary; the chain is the finding. The graph is what lets a tool rank one path above another.

Google documents a narrower version for its recommender. Some role recommendations are associated with lateral movement insights, which identify roles that allow service accounts in one project to impersonate service accounts in another. This is attack path thinking applied to a single relationship, and Google shows it next to the recommendation to remove the role. A path analysis shows these links in sequence, so a reader can decide where to cut it.

## Why ranking by path helps a small team

Teams drown in findings. Microsoft frames the purpose as helping security teams assess the risk behind each issue and identify the highest-risk issues for immediate resolution. A path that starts at the internet and ends at a data store belongs above a hundred isolated misconfigurations. The ranking works because a path shows a route that exists today, not a rule that might be broken.

The ranking is only as good as the graph. A graph with no network model will miss paths that depend on routing and security groups, and a graph with no identity model will miss the hops that matter most. Act argues that network and identity are evaluated by separate teams in separate tools, which is why it models them together. That framing is Act's. Judge any path tool by the parts of the graph it models, and ask what it leaves out.

## A finding is not a fix

The Defender documentation says the algorithm exposes attack paths and suggests recommendations to fix issues, breaking the path. The recommendation is a suggestion, and someone must apply it. Wiz, Tenable and Defender for Cloud present paths as prioritized findings. Applying the change is a separate step in a separate team's queue. The distinction is between knowing a route exists and having closed it.

This is why the index gives 25 percent of the weight to whether a product ends with an enforced change. A path analysis that ends in a ticket helps with prioritization and leaves the removal to people. [Sonrai Security](/platforms/sonrai/) removes access through policies the customer deploys, while [Act](/platforms/act/) delivers changes as pull requests. See [the method](/method/) for how the two were scored. The cut can be a policy, a network rule or a trust change, and each is a change that someone has to author, test and ship.

## How to use paths in a program

Use the paths to choose boundaries. If twenty paths cross the same trust relationship or the same unused region, one boundary removes all of them. Simulate it, enforce it, then rerun the analysis and check that the paths are gone. The rerun is the proof, and it is the step that findings-only tools leave out. The rerun gives a repeatable measure that does not depend on any one vendor's score.

Keep a short list of crown-jewel resources and report the number of paths to each, month by month. A falling number is evidence an auditor can read. A rising number after a quiet quarter is the signal that new access is arriving faster than it is removed. Make the monthly number part of the security review, so the trend is read by someone who can act on it.

## Sources

- Microsoft Learn, attack paths in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/concept-attack-path (read 10 Nov 2026)
- Microsoft Learn, CIEM in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 10 Nov 2026)
- Google Cloud, overview of role recommendations: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 10 Nov 2026)
- Wiz Academy, how to choose a CIEM vendor: https://www.wiz.io/academy/cloud-security/how-to-choose-a-ciem-vendor (read 10 Nov 2026)
- Tenable, enforce least privilege across cloud identities: https://www.tenable.com/cloud-security/use-cases/enforce-least-privilege-across-cloud-identities (read 10 Nov 2026)
- Act blog, network and identity each see half your cloud: https://act.security/resources/network-and-identity-each-see-half-your-cloud-your-attacker-sees-both (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://cloudleastprivilege.com/guides/attack-path-analysis/
