Skip to main content

Pentrova is launching soon. Join the waitlist for early access.Join the waitlist

Research

Attack-chain taxonomy 101: the five classes Pentrova organises coverage around

Pentrova groups attack chains into five classes so teams fix them faster. Here are the classes, why they beat a flat CVSS list, and how each maps to coverage.

Reading mode

A taxonomy is useful when it shapes the work. Pentrova organises attack chains into five classes — Injection, Access Control, Crypto, Infra, and AppSec — and each one exists because it maps cleanly to a coverage area and to a shared review discipline.

This post names the five classes, shows how each one shares a structural shape, and explains why operational clarity matters more than a perfect mapping to CWE.

Why classify chains at all?#

Before this taxonomy existed, findings tended to live as a flat list ranked by severity. A flat list is hard to own: nobody can reason about “all 400 findings” at once, and a severity number tells you nothing about who should fix a bug or how it should be reviewed. Grouping by class gives the work a shape — engineers reason about one class at a time, auditors map one class to one control family, and the catalog grows in blocks rather than as an ever-longer list.

The five classes#

Injection#

Injection chains share a shape: user input reaches an evaluation context. The evaluator differs across SQL, templates, XML, and scripting sinks, but the chain structure and the evidence look similar enough that a single team can own the class. XXE and both live here despite very different payloads.

Access Control#

Access Control chains share a different shape: a caller accesses an object the policy should have refused. The confirmation is always a cross-role replay with a diff. Coverage differs between REST, GraphQL, and session-based apps, but the evidence standard is identical — which is exactly what the Authorization Matrix is built to produce.

Crypto#

Crypto chains turn on weak or misapplied cryptography — predictable tokens, broken signature verification, downgradeable transport. The shape is “a guarantee the system assumed turns out not to hold”.

Infra#

Infra chains exploit the platform beneath the application — exposed metadata services, misconfigured storage, reachable internal endpoints. SSRF reaching cloud metadata is the canonical example.

AppSec#

AppSec is the catch-all for application-specific logic — workflow bypasses, race conditions, business-rule violations. These are the findings only the application can define, the kind covered in the race-condition playbook.

Why operational clarity beats CWE accuracy#

The point of the taxonomy is not accuracy against CWE — the product carries a CWE cross-walk inside for anyone who needs the formal mapping — but operational clarity. The five classes let engineers reason about one class at a time, let auditors map one class to one control family, and let the catalog grow in blocks. The classes let the work move, which is what a taxonomy is for.

This grouping also feeds the escalation catalog: chains compose across classes (an Injection finding escalating through an Infra path), and naming the classes makes those cross-class chains legible.

Key takeaways#

  • Pentrova uses five chain classes: Injection, Access Control, Crypto, Infra, AppSec.
  • Each class shares a structural shape and a shared review discipline, so a team can own it.
  • The taxonomy optimises for operational clarity, not CWE precision; a CWE cross-walk lives in the product.
  • Classes let the catalog grow in blocks and make cross-class escalation chains legible.

FAQ#

Does this taxonomy replace CWE or OWASP categories? No. It is an operational grouping for ownership and review. Pentrova carries a CWE cross-walk internally, so a finding still maps to its formal CWE for compliance and reporting.

Why five classes and not more? Five is enough to give the work a shape without fragmenting ownership. More classes would split teams too finely; fewer would lump unrelated review disciplines together.

How does a chain that spans classes get categorised? By its escalation path. A chain that starts with Injection and pivots through Infra is legible precisely because both classes are named — the catalog records the cross-class sequence.

Read how chains compose in how Pentrova turns single bugs into exploit chains, or explore the docs.

Updated

Written by

Pentrova Engineering Pentrova Research

Pentrova Research writes about deterministic offensive-security proof, LLM-driven pentest chains, and how to ship exploit-grade evidence into engineering pipelines.

Keep reading

Site search

↑↓ navigateEnter openEsc close