Skip to main content
Data Remediation is coming soon. This page describes planned functionality.

​
Data Remediation

Fix it before it leaves the door. Proactive AI-assisted remediation. When risky data patterns appear, DataHarbor can apply protection immediately and alert you to review.

​
How it works

  1. Detect — Schema Analytics identifies a risk
  2. Suggest — DataHarbor recommends a remediation
  3. Apply — Optionally auto-apply with human-in-the-loop approval
  4. Verify — Confirm the fix resolves the issue

​
Remediation modes

​
Manual (default)

Receive alerts, apply fixes yourself: 
remediation:
  mode: manual
  # You receive alerts; you apply controls

​
Suggested

DataHarbor suggests fixes with one-click apply: 
{
  "alert": "uncovered_sensitive_field",
  "field": "social_security_number",
  "suggested_remediation": {
    "control": "redact",
    "confidence": 0.98
  },
  "actions": {
    "apply": "POST /v1/remediations/rem_123/apply",
    "dismiss": "POST /v1/remediations/rem_123/dismiss"
  }
}

​
Auto-protect

Automatically apply controls to high-confidence risks, notify for review: 
remediation:
  mode: auto_protect
  threshold: 0.95  # Only auto-apply when confidence >= 95%
  notify: always
When auto-protect triggers: 
{
  "event": "auto_remediation_applied",
  "field": "credit_card_number",
  "control_applied": "redact",
  "confidence": 0.99,
  "status": "pending_review",
  "review_url": "https://app.dataharbor.dev/remediations/rem_456"
}

​
Remediation types

RiskSuggested remediation
SSN, credit card, passportRedact
Email, phone (analytics use)Tokenize
Email, phone (demo use)Anonymize
Name + DOB + ZIP (combo risk)Generalize ZIP to 3-digit
New unknown PII fieldHold for review

​
Human-in-the-loop

Even in auto-protect mode, you maintain control:
  • Review queue: All auto-applied remediations appear for review
  • Undo: Revert any auto-applied control within 24 hours
  • Tune: Adjust confidence thresholds per field type
  • Exclude: Mark specific fields as “known safe”
remediation:
  mode: auto_protect
  threshold: 0.95
  
  exclusions:
    - field: internal_reference_id
      reason: "Not PII, just looks like one"

​
Audit trail

Every remediation is logged: 
{
  "remediation_id": "rem_789",
  "timestamp": "2025-01-15T10:30:00Z",
  "virtual_api": "partner-view",
  "field": "account_number",
  "risk_detected": "uncovered_sensitive_field",
  "action": "auto_redact",
  "confidence": 0.97,
  "reviewed_by": "jane@yourcompany.com",
  "reviewed_at": "2025-01-15T14:22:00Z",
  "review_decision": "approved"
}

​
Next steps

Schema Analytics

Understand what triggers remediation

Data Control

Manual control application