Show HN: Aethel-Core – Ultra-Lightweight Neuro-Symbolic WAF Built in SWI-Prolog (github.com)

Aethel-Core has unveiled a cutting-edge neuro-symbolic web application firewall (WAF) developed in SWI-Prolog, designed to deliver ultra-lightweight, high-fidelity cyber security. This hybrid system emphasizes real-time threat mitigation and LLM instruction tuning, boasting features like dynamic backtracking for querying threat signatures in RAM and advanced volumetric DDoS protection using sliding time-window tracking. Key innovations include n-gram tokenization to break long signatures into smaller components, and a recursive normalization pipeline that effectively unpacks various encoding formats. This development is noteworthy for the AI/ML community as it demonstrates a significant leap in processing efficiency and security integrity. In rigorous tests against a complex Hyper-Chaos Mutation Attack, Aethel-Core achieved remarkable metrics, processing over 74 million logical inferences and maintaining a 100% security integrity rate with zero bypasses. Its low volatile RAM footprint of approximately 2.7 MB and an impressive peak induction speed of nearly 6.8 million inferences per second highlight its capability for real-time defense against diverse threats, including Layer-7 traffic flooding and adversarial AI attacks. The open-source repository provides insights into its core functionality, even as critical components remain proprietary.