Skip to Content
DetectorsOverview

Detectors

Detectors are the scanning engines that read the content pulled from your sources and flag anything that matters: hardcoded credentials, personal data, malware signatures, insecure code, broken links — or any signal you define yourself. Each thing a detector flags becomes a finding.

There are two kinds, and they work side by side:

What it isBest for
Pre-built detectorsCurated, ready-to-use packs that ship with ClassifyreCommon, universal signals — turn them on and go
Custom detectorsDetectors you build, from a simple rule to an AI modelSignals specific to your team, domain, or policy

The Source → Asset → Finding model

Everything detectors do sits inside one simple hierarchy:

  1. Source — a connection to a system you run.
  2. Asset — one item extracted from it (a page, file, table, message, video), carrying content and metadata.
  3. Finding — one signal a detector raised on one asset. An asset can produce zero, one, or many findings.

A scan ingests a source into assets, each asset’s content is read by the detectors configured on that source, and the result is a list of findings you can triage and investigate.


Start here

PageWhat you’ll learn
How Detectors WorkHow detectors run, how they’re matched to content, and how you switch them on per source.
Findings & ResultsWhat a detector produces — every field of a finding, severity levels, confidence, and the status lifecycle.
Pre-built DetectorsThe ready-made packs and their configuration.
Custom DetectorsBuild your own — from a regex to an LLM, across text and images.

Pre-built detectors

Classifyre ships with ready-made detectors for the most common signal types, so you get findings on day one with nothing to train:

  • Secrets — hardcoded credentials, API keys, tokens
  • PII — personally identifiable information
  • YARA — malware patterns and threat indicators
  • Broken Links — unreachable or empty URLs
  • Code Security — insecure code patterns

See Pre-built Detectors for the full catalog and configuration options.


Custom detectors

When a signal is specific to you, build a custom detector. The same framework spans the whole spectrum — from a zero-overhead pattern rule to a full AI model, across both text and images:

EngineModalityBest for
RegexTextCodes, IDs, structured patterns — exact and instant
GLiNER2TextEntity extraction with no labelled training data
AI Detector (LLM)TextNuanced classification and structured extraction
Text ClassificationTextSpam, toxicity, sentiment, topic labels
Image ClassificationImageNSFW, moderation, custom image categories
Feature ExtractionTextEmbeddings for semantic search and clustering
Object DetectionImageLocating objects, label-based severity

See Custom Detectors to pick an engine and build one. The Detector agent can even author and test custom detectors for you automatically.

Last updated on