test-types.md

Chaos Test Types

✅ All test suites passing (24 tests across 5 suites)

Glitch Gremlin AI supports several types of chaos testing to help identify different categories of vulnerabilities and issues in your Solana programs. All test types have been validated through our comprehensive test suite.

Available Test Types

1. Fuzz Testing (FUZZ)

Automatically generates random or semi-random inputs to program instructions to find edge cases and unexpected behaviors.

const request = await sdk.createChaosRequest({
    targetProgram: "Your program ID",
    testType: "FUZZ",
    duration: 300,
    intensity: 5,
    params: {
        instructionTypes: ["all"], // or specific instructions
        seedRange: [0, 1000000]
    }
});

2. Load Testing (LOAD)

Simulates high transaction volume to identify performance bottlenecks and concurrency issues.

const request = await sdk.createChaosRequest({
    targetProgram: "Your program ID",
    testType: "LOAD",
    duration: 600,
    intensity: 8,
    params: {
        tps: 5000,
        rampUp: true
    }
});

3. Exploit Testing (EXPLOIT)

Attempts known exploit patterns to verify program security.

const request = await sdk.createChaosRequest({
    targetProgram: "Your program ID",
    testType: "EXPLOIT",
    duration: 300,
    intensity: 7,
    params: {
        categories: ["reentrancy", "arithmetic"]
    }
});

4. Concurrency Testing (CONCURRENCY)

Tests program behavior under parallel transaction scenarios.

const request = await sdk.createChaosRequest({
    targetProgram: "Your program ID",
    testType: "CONCURRENCY",
    duration: 300,
    intensity: 6,
    params: {
        parallel: 10,
        conflictRate: 0.5
    }
});

Test Parameters

Common Parameters

• duration: Test duration in seconds
• intensity: Scale of 1-10, affects how aggressive the testing is
• params: Optional test-specific parameters
• mlEnabled: Enable AI-driven vulnerability detection (default: true)
• mlConfidenceThreshold: Minimum confidence score for ML predictions (default: 0.7)

ML Model Parameters

The AI engine uses a neural network model to detect potential vulnerabilities:

const request = await sdk.createChaosRequest({
    targetProgram: "Your program ID",
    testType: TestType.EXPLOIT,
    duration: 300,
    intensity: 7,
    params: {
        mlConfig: {
            confidenceThreshold: 0.8,
            featureExtraction: {
                includeStaticAnalysis: true,
                includeDynamicTraces: true
            }
        }
    }
});

Test-Specific Parameters

Each test type accepts specific parameters to customize the testing behavior. See the examples above for type-specific parameters.

Best Practices

1. Start with lower intensity values (1-3) and gradually increase
2. Run tests on devnet first before mainnet
3. Monitor program logs during testing
4. Review all test results carefully, even if no errors are reported

Result Analysis

Test results include:

• Transaction statistics
• Error rates and types
• Performance metrics
• Specific vulnerability findings (if any)
• Recommendations for improvements

Next Steps

• Learn about automated testing integration
• Explore governance features
• Set up monitoring