Astraea Unified Threat: Security Logs and Emails for Multi-Task Model Fine-Tuning

Auren Research's Astraea Unified Threat (v4) is a production-balanced telemetry dataset designed for fine-tuning multi-task encoder architectures. Each example is a structured security log or email payload prefixed with task-specific labels like [TASK=LOG_CLS]. The dataset was last updated on May 30, 2026.

Use Cases

• Log triage based on structured security log entries.
• Indicator of Compromise (IOC) extraction from email payloads.
• MITRE ATT&CK technique mapping based on described attack patterns.
• Attack-chain classification using the task-specific prefixes.

Strengths

• Designed for production-balancing, which suggests consideration for real-world application.
• Includes task-specific prefixes enabling multi-task learning for at least five distinct cybersecurity tasks.
• Last updated on May 30, 2026, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to the original telemetry collection.

Provenance

Source

Auren Research

Collection Method

Likely collected from security telemetry systems and email payloads.

Freshness

Last updated 2026-05-30 20:09:23