Skip to content

Weaviate (ailab-ds)

What It Is

Weaviate is an open-source vector database that supports semantic search and hybrid queries. On the data science workstation, it runs as a standalone binary with anonymous access explicitly enabled — no authentication required. Three classes store a mix of sensitive research documents, team communications, and benign public content.

Installation

Installed via binary download from GitHub releases:

wget https://github.com/weaviate/weaviate/releases/download/v1.28.4/weaviate-v1.28.4-linux-amd64.tar.gz
tar xzf weaviate-v1.28.4-linux-amd64.tar.gz
mv weaviate /usr/local/bin/weaviate
chmod +x /usr/local/bin/weaviate

Release availability

Verify the release URL at build time. Binary releases may not be available for all versions or architectures on the Weaviate GitHub releases page.

Systemd Unit

Unit: weaviate.service

/etc/systemd/system/weaviate.service
[Service]
User=root
ExecStart=/usr/local/bin/weaviate --host 0.0.0.0 --port 8080 --scheme http
Environment="PERSISTENCE_DATA_PATH=/opt/weaviate/data"
Environment="AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true"
Environment="QUERY_DEFAULTS_LIMIT=25"
Environment="DEFAULT_VECTORIZER_MODULE=none"
Restart=always

Misconfiguration

AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true explicitly disables authentication. Combined with binding to 0.0.0.0, all classes and objects are accessible to anyone on the network.

Port & Config

Parameter Value
Host 172.16.50.30
Port 8080
Bind address 0.0.0.0
Authentication Anonymous access enabled
Data directory /opt/weaviate/data/
Vectorizer None (pre-computed vectors)

Seed Data

Three classes are seeded:

Class Objects Content
ResearchDocument 6 Database credentials, API keys embedded in internal research notes
TeamCommunication 6 SSNs, salary information in team chat logs and HR communications
PublicKnowledge varies Benign public-facing content (noise)

What aipostex Finds

  • Unauthenticated access — The Weaviate API requires no credentials, confirmed by the AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true environment variable in the systemd unit.
  • Class enumeration — The /v1/schema endpoint lists all classes and their properties, revealing the data model.
  • Sensitive data extraction — Querying ResearchDocument and TeamCommunication surfaces database credentials, API keys, SSNs, and salary data.

Verification

Confirm Weaviate is running:

curl http://localhost:8080/v1/meta
Expected response 
{"hostname": "...", "version": "1.28.4", ...}

List schema classes:

curl http://localhost:8080/v1/schema
Expected response

JSON listing ResearchDocument, TeamCommunication, and PublicKnowledge classes with their property definitions.