# Oracle Deep Data Security Lab

Internal enablement kit for demonstrating Oracle Deep Data Security in realistic enterprise data security scenarios: AI agents, prompt injection, shared application accounts, ad hoc BI, PII access, view bypass, legacy app modernization, RAG/vector retrieval, and audit evidence.

The goal is to let any team member deploy a secure OCI foundation, install lab scenarios, run positive and negative tests, and collect evidence in a repeatable way.

## Quick View

```text
terraform/   Secure-by-default OCI infrastructure
scenarios/   Independent SQL labs and runbooks
scripts/     Bootstrap, validation, execution, and reset automation
docs/        Architecture, demo, and operations guides
apps/        Placeholder for Spring Boot app, AI agent, and BI simulator
```

## Included Scenarios

| ID | Scenario | Objective |
| --- | --- | --- |
| 01 | AI Prompt Injection | Show that the database limits data even when an AI agent generates broad or malicious SQL. |
| 02 | Shared App Account | Show end-user data enforcement even when a technical application account is used. |
| 03 | PII Row/Column/Cell | Show row, column, and cell-level controls for personal data and salary. |
| 04 | View Bypass / MAC | Show Mandatory Access Control with `USE DATA GRANTS ONLY`. |
| 05 | Legacy App AI Extension | Show AI modernization without rewriting all legacy application authorization logic. |
| 06 | RAG Vector Classified Docs | Show RAG/vector search returning only authorized chunks by classification. |
| 07 | Audit Evidence With Data Safe | Show access evidence with Unified Audit and OCI Data Safe validation guidance. |

Each scenario includes a `RUNBOOK.md` with a before/after demo flow, expected evidence, and official references.

## Scenario Execution Guides

Use these links to open the step-by-step demo guide for each scenario. Each runbook explains the vulnerable baseline, the commands to apply Oracle Deep Data Security, the expected protected result, and the official references.

| ID | Scenario | Step-by-step guide |
| --- | --- | --- |
| 01 | AI Prompt Injection | [RUNBOOK.md](scenarios/01-ai-prompt-injection/RUNBOOK.md) |
| 02 | Shared App Account | [RUNBOOK.md](scenarios/02-shared-app-account/RUNBOOK.md) |
| 03 | PII Row/Column/Cell | [RUNBOOK.md](scenarios/03-pii-row-column-cell/RUNBOOK.md) |
| 04 | View Bypass / MAC | [RUNBOOK.md](scenarios/04-view-bypass-mac/RUNBOOK.md) |
| 05 | Legacy App AI Extension | [RUNBOOK.md](scenarios/05-legacy-app-ai-extension/RUNBOOK.md) |
| 06 | RAG Vector Classified Docs | [RUNBOOK.md](scenarios/06-rag-vector-classified-docs/RUNBOOK.md) |
| 07 | Audit Evidence With Data Safe | [RUNBOOK.md](scenarios/07-audit-evidence-data-safe/RUNBOOK.md) |

## Prerequisites

- OCI tenancy with permission to create networking, Autonomous Database, optional Vault, and optional Compute resources.
- Terraform 1.6 or later.
- OCI CLI configured, or OCI provider authentication variables.
- SQLcl, SQL*Plus, or another compatible Oracle client.
- Access to an Oracle AI Database version compatible with Oracle Deep Data Security.

## 7-Step Execution

1. Clone the repository.

2. Copy the example Terraform variables file.

   Linux/macOS:

   ```bash
   cp terraform/envs/demo/terraform.tfvars.example terraform/envs/demo/terraform.tfvars
   ```

   Windows PowerShell:

   ```powershell
   Copy-Item terraform\envs\demo\terraform.tfvars.example terraform\envs\demo\terraform.tfvars
   ```

3. Edit `terraform/envs/demo/terraform.tfvars` with your OCIDs, region, and database parameters.

   Linux/macOS:

   ```bash
   vi terraform/envs/demo/terraform.tfvars
   ```

   Windows PowerShell:

   ```powershell
   notepad terraform\envs\demo\terraform.tfvars
   ```

4. Validate the infrastructure.

   Linux/macOS:

   ```bash
   chmod +x scripts/*.sh
   ./scripts/validate-terraform.sh
   ```

   Windows PowerShell:

   ```powershell
   powershell -ExecutionPolicy Bypass -File .\scripts\validate-terraform.ps1
   ```

5. Deploy the infrastructure.

   Linux/macOS:

   ```bash
   cd terraform/envs/demo
   terraform init
   terraform plan -out tfplan
   terraform apply tfplan
   cd ../../..
   ```

   Windows PowerShell:

   ```powershell
   Set-Location terraform\envs\demo
   terraform init
   terraform plan -out tfplan
   terraform apply tfplan
   Set-Location ..\..\..
   ```

6. Install a scenario.

   Linux/macOS:

   ```bash
   ./scripts/run-scenario.sh 01-ai-prompt-injection "<connect_string>"
   ```

   Windows PowerShell:

   ```powershell
   powershell -ExecutionPolicy Bypass -File .\scripts\run-scenario.ps1 -Scenario 01-ai-prompt-injection -ConnectString "<connect_string>"
   ```

7. Run tests and reset when needed.

   Linux/macOS:

   ```bash
   ./scripts/run-scenario.sh 05-legacy-app-ai-extension "<connect_string>"
   ./scripts/reset-scenario.sh 05-legacy-app-ai-extension "<connect_string>"
   ```

   Windows PowerShell:

   ```powershell
   powershell -ExecutionPolicy Bypass -File .\scripts\run-scenario.ps1 -Scenario 05-legacy-app-ai-extension -ConnectString "<connect_string>"
   powershell -ExecutionPolicy Bypass -File .\scripts\reset-scenario.ps1 -Scenario 05-legacy-app-ai-extension -ConnectString "<connect_string>"
   ```

## Secure Defaults

- Database deployed in a private subnet.
- No public IP on the database.
- Dedicated NSGs for application and database access.
- mTLS required for Autonomous Database connectivity.
- Secrets kept out of Git.
- Optional Vault/KMS customer-managed keys.
- Compute bastion disabled by default.
- Demo evidence and logs ignored by Git.

## Contributing

Read [CONTRIBUTING.md](CONTRIBUTING.md). Every new scenario must include `README.md`, `RUNBOOK.md`, `metadata.yaml`, numbered SQL files, positive/negative tests, and a reset script.

## CI/CD

This repository includes GitHub Actions for:

- `terraform fmt`
- `terraform init -backend=false`
- `terraform validate`
- minimum scenario structure checks
- blocking sensitive files such as `.tfvars`, `.pem`, and `.key`