ai-assurance

This repository serves as a comprehensive resource for integrating machine learning with security operations, offering innovative cybersecurity strategies. It features a mix of practical code examples, insightful research, and valuable resources tailored for advancing AI/ML cyber security practices.

A neural network verification tool based on the DPLL(T) SMT Solving algorithm.

Scan your AI/ML models for problems before you put them into production.

Local-first LLM adversarial assurance lab for mapping attacks to controls, evidence, and framework readiness.

Trusted autonomy T&E runtime that links mission needs, hazards, scenarios, telemetry, evidence, verification reports, and hash-chained ledgers so AI/autonomous decisions can be reviewed instead of merely trusted.

Proof-to-pilot AI validation architecture with reproducible benchmarks, cryptographic evidence custody, and fail-closed review gates.

Assurance methodology for autonomous AI agents in regulated financial services

Local-first coding agent verification and adversarial review with frozen evidence, moderated findings, and human final acceptance.

ICSThreatQA: A human-in-the-loop benchmark for evaluating knowledge-grounded QA, RAG, and AI assurance in Industrial Control System cybersecurity.

Local-first AI coding assistant with 9-layer security, adversarial model testing, and cryptographic audit trails. Break your AI before it breaks your code.

TrustLens — explainable & assured AI for Zero Trust continuous authentication. Verify the explanation itself: cross-method XAI consensus, adversarial explanation-robustness auditing & PSI concept-drift monitoring. SHAP/LIME/Anchor, STRIDE threat model, NIST SP 800-207, Streamlit dashboard. By Krishita Sanjay Choksi.

Evidence-bound AI assurance framework for dependence-aware guardrail composition, scoped safety claims, reproducible reports, and receipt-backed disclosure controls.

Open documentary assurance method for evaluating whether human authority retained practical force in AI-mediated institutional decisions.

Governed cognition substrate for AI engineering agents: mission envelopes, evidence-bound belief/plan graphs, quarantined memory, model-role review, sentinel checks, and BlackFox handoffs under human authority.

Source-available governed cognitive architecture with an IX language runtime, typed VM, memory, planning, world modeling, uncertainty, learning, evidence-gated proposals, staged execution, human review, audited reentry, closeout ledgers, and exportable receipts. AI proposes. Humans decide. No AGI claims.

Policy-as-code for AI engineering evidence. Verify AI controls, evals, exceptions, and risk policy in CI.

Evidence architecture for justified reliance and practical human control in AI-mediated decisions.

Artificial Intelligence Assurance Protocol (AIAP) — an auditable framework for governing AI use through competence, independence, disclosure, verification, and accountable human judgment.

Machine-checked bridge from accepted native TrackB replay witnesses to formal replay-obstruction conclusions.

Professional AI Security Assurance portfolio demonstrating model supply-chain security, LLM red teaming, static analysis, SBOM validation, risk classification, and governance-aligned AI safety workflows.