This project provides a stable, standardized baseline for participating labs to contribute AI security tools in a consistent and reusable manner. Our goal is to secure the entire AI lifecycle—from training data inspection to agentic deployment.
- Standardized Specifications: Definition of directory structures and JSON file contracts to ensure seamless tool interoperability.
- Security Tools: Specialized toolsets developed by various labs, covering data sanitization, vulnerability detection, and attack mitigation.
- AIBOM Integration: Automated generation and mapping of AI Bills of Materials (AIBOM) for comprehensive risk assessment.
- E2E Demo: A lightweight, Docker-first workflow designed to test the entire integrated pipeline with minimal setup.
- ai-lifecycle-spec: Definitions for standard directory structures, JSON schemas, and example I/O payloads.
- ai-lifecycle-demo: End-to-end runnable demo environment (Docker-based).
Each tool adheres to the project’s standard directory and JSON specifications while focusing on a specific security domain.
| Repository Name | Description | Research Organization |
|---|---|---|
| T1-Data-Sanitization | Training Data Poisoning Inspection & Sanitization Report. | SKKU |
| T2-AI-Model-Vulnerability-Detection | Post-Training Model Vulnerability & Backdoor Detection. | SKKU |
| T3-AI-Agent-Vulnerability-Detection | Scenario-Based Vulnerability Detection for Agentic AI Systems. | SKKU |
| T4-AIBOM-Generation | Automated AIBOM Generation (Model, Library, and Dependency Scanning). | Korea Univ. |
| T5-AIBOM-based-Vulnerability-Detection | AIBOM-Based Vulnerability Mapping & Risk Impact Assessment. | Korea Univ. |
| T6-Attack-Mitigation | Feasibility Assessment of Countermeasure Policies by Attack Scenario. | Korea Univ. |
| TTA-repo | Repository for TTA | TTA |
| MCCAAi-repo | Repository for MCCAAi | MCCAAi |
| CSSA-repo | Repository for CSSA | CSSA |
- Understand the Specs: Start with ai-lifecycle-spec to review the required directory structures and JSON file formats.
- Run the Demo: Use ai-lifecycle-demo to experience how all tools integrate based on the shared specifications.
- Explore Tools: Visit individual repositories for specific implementation details and technical stacks.
We welcome contributions from participating labs and external collaborators. To ensure efficient collaboration:
- Please keep pull requests small and focused.
- Include tests whenever applicable.
- Ensure all tool outputs remain fully compatible with the defined directory and JSON specs.
- For major design proposals, please open an issue or discussion first.
- For technical discussions, please use GitHub Issues/Discussions in the relevant repository.
- For cross-lab coordination, please reach out to the designated contacts listed in the project notes.
Unless otherwise stated, repositories in this organization are released under the license specified within each individual repository.
