Noise Configuration¶

Configure noise filtering to reduce false positives and focus on actionable findings.

Overview¶

Noise configuration controls how SuperQode filters and deduplicates QE findings:

Confidence Thresholds: Filter low-confidence findings
Deduplication: Remove similar findings
Severity Filtering: Focus on important issues
Known Risk Suppression: Suppress acknowledged risks
Limits: Control finding volume

Basic Configuration¶

qe:
  noise:
    min_confidence: 0.7
    deduplicate: true
    min_severity: "low"

Configuration Options¶

Confidence Threshold¶

Minimum confidence score (0.0 to 1.0) for findings:

qe:
  noise:
    min_confidence: 0.7  # Default: 0.7

Recommendations:

Use Case	Threshold	Rationale
Strict filtering	`0.9`	Only high-confidence findings
Balanced (default)	`0.7`	Good balance of precision/recall
More findings	`0.5`	Include more potential issues

Deduplication¶

Remove similar findings:

qe:
  noise:
    deduplicate: true  # Default: true
    similarity_threshold: 0.8  # Default: 0.8

Similarity Threshold: 0.0 (different) to 1.0 (identical)

0.8: Default, removes clearly duplicate findings
0.9: Only removes nearly identical findings
0.7: More aggressive deduplication

Severity Filtering¶

Minimum severity to report:

qe:
  noise:
    min_severity: "low"  # low, medium, high, critical

Severity Levels:

critical: Only critical issues
high: Critical and high severity
medium: Critical, high, and medium
low: All findings (default)

Known Risk Suppression¶

Suppress findings matching known risk patterns:

qe:
  noise:
    suppress_known_risks: true
    known_risk_patterns:
      - "Deprecated API usage"
      - "TODO:.*security"
      - "FIXME:.*vulnerability"

Pattern Types:

Exact match: Exact string match
Regex patterns: Regular expression patterns (Python regex)

Finding Limits¶

Control maximum number of findings:

qe:
  noise:
    max_findings_per_file: 10   # 0 = unlimited
    max_total_findings: 100      # 0 = unlimited

Use Cases:

Per-file limit: Prevent one file from dominating report
Total limit: Cap total findings for review

Complete Configuration Example¶

qe:
  noise:
    # Confidence threshold
    min_confidence: 0.75

    # Deduplication
    deduplicate: true
    similarity_threshold: 0.85

    # Severity filtering
    min_severity: "medium"

    # Known risk suppression
    suppress_known_risks: true
    known_risk_patterns:
      - "Known issue:.*"
      - "Acceptable risk:.*"
      - "TODO.*security.*acknowledged"

    # Finding limits
    max_findings_per_file: 5
    max_total_findings: 50

    # Severity rules
    apply_severity_rules: true

Severity Rules¶

Enable rule-based severity adjustments:

qe:
  noise:
    apply_severity_rules: true  # Default: true
    severity_rules:
      # Custom severity rules (optional)
      rules:
        - pattern: "SQL.*injection"
          severity: "critical"
        - pattern: "XSS.*vulnerability"
          severity: "high"

Default Rules:

SuperQode includes default severity rules for common issues: - SQL injection → critical - XSS vulnerabilities → high - Hardcoded secrets → critical - And more...

Custom Rules:

Define custom patterns:

qe:
  noise:
    severity_rules:
      rules:
        - pattern: "authentication.*bypass"
          severity: "critical"
        - pattern: "rate.*limit.*missing"
          severity: "medium"

Memory-Based Suppression¶

SuperQode learns from user feedback and suppresses findings that were previously dismissed:

qe:
  noise:
    # Memory-based suppression is automatically enabled
    # when memory.store is configured
    use_memory_suppressions: true  # Default: true if memory enabled

How It Works:

User dismisses a finding
Finding pattern stored in memory
Future similar findings automatically suppressed

Configuration:

memory:
  store:
    type: file
    path: .superqode/memory

Mode-Specific Configuration¶

Different noise settings for Quick Scan vs Deep QE:

qe:
  modes:
    quick_scan:
      noise:
        min_confidence: 0.8
        min_severity: "medium"
        max_total_findings: 20

    deep_qe:
      noise:
        min_confidence: 0.6
        min_severity: "low"
        max_total_findings: 200

Rationale:

Quick Scan: Stricter filtering for fast feedback
Deep QE: More permissive to catch edge cases

Advanced Options¶

Similarity Calculation¶

Deduplication uses multiple factors:

File path
Line number (nearby lines considered similar)
Finding title and description
Evidence content

Custom Similarity:

qe:
  noise:
    similarity_threshold: 0.8
    # Uses SequenceMatcher for text similarity

Severity Adjustment Logging¶

Log severity adjustments for debugging:

qe:
  noise:
    apply_severity_rules: true
    # Severity adjustments logged at DEBUG level

Configuration Priority¶

Noise configuration is loaded in this order:

Role-specific (in team.modes.qe.roles.<role>.noise)
Mode-specific (in qe.modes.<mode>.noise)
Global (in qe.noise)

Example:

qe:
  noise:
    min_confidence: 0.7  # Global default

  modes:
    quick_scan:
      noise:
        min_confidence: 0.8  # Override for quick scan

team:
  modes:
    qe:
      roles:
        security_tester:
          noise:
            min_confidence: 0.9  # Stricter for security role

Testing Configuration¶

Test noise configuration:

# Run QE with verbose output
superqe run . --mode quick -v

# Check filtering stats in QR
cat .superqode/qe-artifacts/qr-*.json | jq '.noise_filter_stats'

Best Practices¶

1. Start Conservative¶

Begin with stricter filtering, then relax:

qe:
  noise:
    min_confidence: 0.8
    min_severity: "medium"

2. Use Known Risk Patterns¶

Suppress acknowledged risks:

qe:
  noise:
    suppress_known_risks: true
    known_risk_patterns:
      - "Acknowledged.*TODO"
      - "Acceptable.*risk"

3. Set Reasonable Limits¶

Prevent report bloat:

qe:
  noise:
    max_findings_per_file: 10
    max_total_findings: 100

4. Adjust by Mode¶

Stricter for Quick Scan, permissive for Deep QE:

qe:
  modes:
    quick_scan:
      noise:
        min_confidence: 0.8
        max_total_findings: 20
    deep_qe:
      noise:
        min_confidence: 0.6
        max_total_findings: 200

Troubleshooting¶

Too Many Findings¶

Problem: Report has too many findings

Solution: Increase thresholds and limits:

qe:
  noise:
    min_confidence: 0.8
    min_severity: "medium"
    max_total_findings: 50

Missing Important Findings¶

Problem: Important findings filtered out

Solution: Lower thresholds:

qe:
  noise:
    min_confidence: 0.6
    min_severity: "low"

Duplicate Findings¶

Problem: Same finding appears multiple times

Solution: Enable deduplication:

qe:
  noise:
    deduplicate: true
    similarity_threshold: 0.85  # Higher = more aggressive

Next Steps¶

Noise Filtering - How noise filtering works
YAML Reference - Complete configuration reference
Memory System - Memory-based suppression