LICITRA Evidence Repository

Cryptographically Verifiable Runtime Integrity for Agentic AI Systems.

This repository contains reproducible experiments, exported evidence bundles, and supporting documentation for LICITRA.

---

Quick Start (5 Minutes)

1. Run LICITRA core locally (see docs/how_to_run.md)
2. Follow the demo walkthrough in demo/README.md
3. Inspect generated JSON + PDF evidence artifacts

👉 Start here: demo/README.md

---

What’s Included

• docs/ - threat model, scope, and experiment methodology
• demo/ - reproducible demo walkthrough + exported evidence bundles (JSON/PDF)
• docs/experiments.md - step-by-step commands to reproduce tamper detection tests
• docs/proof.md - summary of results and what each artifact demonstrates
• docs/how_to_run.md - instructions to run the system locally

---

Evidence Summary

Included demo artifacts demonstrate:

• org1: intentional tampering detected at bad_index
• org2: remains clean (multi-organization isolation)
• JSON + PDF forensic exports
• Deterministic SHA-256 hash-chain verification

Artifacts:

• demo/evidence_org1_tampered.json
• demo/evidence_org1_tampered.pdf
• demo/evidence_org2_ok.json
• demo/evidence_org2_ok.pdf

---

Key Claims Demonstrated

• Hash-chained semantic event integrity (tamper detection)
• Multi-organization isolation (org1 chain independent from org2)
• Detection of record mutation, deletion, and previous-hash corruption
• Evidence bundles suitable for governance / audit workflows

---

Architecture

Core implementation lives here:

https://github.com/narendrakumarnutalapati/licitra-core

The core repository contains:

• FastAPI backend
• PostgreSQL ledger implementation
• Deterministic SHA-256 chaining
• Tamper simulation endpoints
• Evidence export (JSON + PDF)

---

Version

Evidence Bundle Version: 0.1-evidence

---

LICITRA provides cryptographic runtime integrity - not observability.

Every AI interaction becomes a chained, verifiable event.

Any historical modification breaks the chain and is detected immediately.

This is forensic-grade runtime governance for agentic AI systems.