Content scanners (CS)

October 31, 2026

Taxonomies › Email infrastructure › Content scanners

Layer 07 of 09 · 13 threats · prefix CS

A scanner is a probabilistic control that an attacker can study, evade and attack directly. These entries cover URL analysis, malware detection, sandboxes, machine learning (ML) and large language model (LLM) classifiers, and data loss prevention (DLP).

Scope: URL, malware, sandbox, ML and LLM inspection engines.

Table: Content scanners (CS), 13 threats with their mitigation principles and design measures. The full text is in the PDF and Excel downloads on the email infrastructure taxonomy overview page.

Text version: every entry on this page is in the PDF and Excel downloads on the overview page, where it can be searched and copied.

The threats

CS·A Analysis-Aware Content
CS·K Scanner Prompt Injection
CS·Q Gated Content
CS·J Reputation Laundering
CS·10 Classifier Evasion
CS·9 Classifier Poisoning
CS·8 Scanner as Attack Surface
CS·7 Scanner Side Effects
CS·6 Verdict Oracle
CS·5 Novelty Window
CS·4 Modality Gap
CS·3 Scanned Data Exposure
CS·2 Exemption Abuse

Where this layer touches others

Delivery pipeline and content scanners: the pipeline decides what to do with a scanner’s limits and verdicts. The scanner layer asks whether the inspection itself holds up.

STRIDE, CAPEC and CWE

Of the 13 entries here, STRIDE fits 5 cleanly and 8 only loosely. CAPEC 3.9 has a pattern that describes 1 of them directly, only a generic parent pattern for 5, and nothing for 7. CWE 4.20 names the weakness directly for 7, has only a broader weakness for 5, and nothing for 1. The table below gives the mapping entry by entry, with the names of the CAPEC patterns and CWE weaknesses; what the mapping shows across the whole taxonomy is in Mapping the Email Threat Taxonomy to STRIDE, CAPEC and CWE. The Excel, SKOS and STIX downloads carry the same mapping as text.

← DNS layer (DN) · Email infrastructure overview · Human-facing layer (HF) →