Skip to content
Tutorials Center
My learning

Topics

Lesson 9 of 38 ยท Shared Responsibility and Threats

Attack Paths and Blast Radius

Learn how a single weakness chains into a larger cloud compromise, how to measure blast radius, and how account separation and permissions limit it, with a runnable graph example.

A single weakness is rarely the whole story of a breach. An attacker enters through one small hole, then looks for what that foothold can reach, and keeps going. Defenders describe this as an attack path: a chain of steps from an entry point to something valuable. The blast radius is how much damage a single compromised identity, workload or account can do. Good cloud design keeps blast radius small and attack paths short on the attacker's side and long on the defender's.

Thinking in graphs

A cloud environment is a graph. Identities, resources and networks are nodes; permissions, trust relationships and network routes are edges. "Identity A can assume role B", "role B can read bucket C", "function D runs as role B", "server E is reachable from the internet". Walk the edges from an exposed node and you see everything an attacker could reach.

flowchart LR
    Net["Internet"] -->|"exposed port"| Web["Web server"]
    Web -->|"runs as"| RoleW["Web role"]
    RoleW -->|"can assume"| RoleA["Admin role"]
    RoleA -->|"full access"| Data[("Customer data")]
    RoleW -->|"read"| Logs[("Log bucket")]

One exposed web server leads to customer data because the web role is allowed to assume an admin role. Remove that one edge and the path disappears.

Anatomy of a chained compromise

Here is a realistic, defender-oriented sequence at a high level:

  1. An application has a flaw that lets an outsider make the server fetch internal URLs.
  2. The server can reach a metadata service that hands out temporary credentials for its role.
  3. The role is far broader than the application needs, so the credentials can list and read many storage buckets.
  4. Logging is not centralized, so the activity looks like normal application traffic.

Each of these is individually fixable: validate outbound requests, require a hardened metadata version, narrow the role, and send logs to a protected place. The lesson is that fixing any one link breaks the chain, and fixing several makes success unlikely.

Measuring blast radius

For a given starting identity, ask: which resources can it reach through direct permissions and through chains of role assumption? The count (and the sensitivity) of reachable resources is its blast radius. Typical reducers:

  • Account separation. Use separate accounts or subscriptions for production, development, security tooling and logging. A credential stolen in development should not reach production data.
  • Narrow roles. One role per workload with only the actions and resources it needs.
  • Permission boundaries and service control policies. Organization-level limits that even an admin of a child account cannot exceed.
  • Network segmentation. Private subnets and tight rules so a compromised server cannot talk to everything.
  • Short-lived credentials. Stolen tokens expire.
  • Regional and service restrictions. Deny use of regions and services you do not need.

Shrink the blast radius

Separate accounts Prod, dev, logs
Narrow roles One per workload
Segment networks No flat networks
Short lifetimes Tokens expire

Assume one thing will be breached; make that thing small.

1 / 2

Runnable example: reachability in a permission graph

The program below models the graph from the diagram, then computes everything reachable from the internet-facing server and finds the shortest path to the customer data. Then it removes one risky edge and checks again. This is the same idea as commercial tools that map attack paths, just at toy scale.

Python Compute blast radius and find a path

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.

After the fix, the only thing reachable from the exposed server is the log bucket, and there is no path to customer data. Removing a single over-broad permission cut the blast radius dramatically.

Defender workflow

  1. Inventory identities and relationships. Users, roles, service accounts, trust policies, network routes.
  2. Compute paths from entry points to crown jewels. Entry points are anything exposed to the internet or to less-trusted accounts; crown jewels are your most sensitive data and keys.
  3. Break the cheapest links first. Narrowing a role is usually cheaper than rebuilding a network.
  4. Add detection on the remaining paths. Alert when an identity does something off its normal path, such as assuming a rarely used role.
  5. Rehearse. Run tabletop exercises: "this web server is compromised, what can it reach?"

Authorized testing only. Real attack-path analysis tools read configuration with read-only permissions from accounts you own. Red-team style exercises on production need written scope and approval.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What is blast radius?

    Blast radius measures what a single compromise can reach, so smaller is better.
  2. 2. Why does removing one over-broad permission often stop a whole attack path?

    Attack paths depend on each step being possible, so removing one edge can disconnect the target.
  3. 3. Which design choice best limits the impact of a stolen development credential?

    Account boundaries stop credentials from one environment reaching another.
  4. 4. In the graph model, what are edges?

    Edges express what can reach or act on what, which is what an attacker follows.

Summary

  • Breaches are usually chains of small weaknesses; the chain is an attack path.
  • Model the environment as a graph of identities, resources and routes to see what each foothold can reach.
  • Reduce blast radius with account separation, narrow roles, boundaries, segmentation and short-lived credentials.
  • Break the cheapest links first and add detection on the paths that remain.

Next: Identity and Access Management Fundamentals