LanG 2.0 enables energy-efficient agentic AI for security operations
Original title: LanG 2.0: Energy-Efficient Agentic AI for Semi-Automated Security Operations at Scale
LanG 2.0 improves campaign correlation to 51.81% and uses small local models at 650 times less energy than Claude Opus 5.5.
Security operations centers process terabytes of alerts and logs. LanG 2.0 is an agentic system that detects threats, correlates evidence and writes rules using small local models with structural safeguards. It incorporates entity-anchored correlation, human approval of rule changes and a tamper-evident ledger. Tests on eight flow datasets show precision rising from 18.75% to 51.81% and rule acceptance from 60% to 98.33%.
- Campaign correlation precision rises from 18.75% to 51.81%
- Rule acceptance rises from at most 60% to 98.33%
- 73.42% F1 on unseen networks at 650 times less energy
- Processes 3.33 million records in 6.4 minutes
- Resists prompt injection, label poisoning and ledger tampering
From the original article
1 min read at arxiv.orgSecurity operations centers (SOCs) need to turn terabytes of alerts, logs and flows into trusted, affordable decisions to unveil hidden threats. We present LanG 2. 0, an agentic system that detects threats, correlates evidence and writes rules, investigating whether small local models can replace frontier models affordably and safely when structural safeguards, not model accuracy, provide the guarantees.
By Anes Abdennebi · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.