data-classification（数据分类）

Data Classification Skill

Purpose

This skill implements data classification requirements as defined in the Hack23 ISMS Data Classification Policy. It ensures that all data is properly classified and handled according to its sensitivity level.

Rules

Classification Levels

Hack23 uses four classification levels:

1. PUBLIC - Information intended for public disclosure
2. INTERNAL - Information for internal use only
3. CONFIDENTIAL - Sensitive information requiring protection
4. RESTRICTED - Highly sensitive information requiring maximum protection

PUBLIC Data

Definition: Information that can be freely disclosed to the public without harm to Hack23, customers, or partners.

Examples:

• Published website content
• Marketing materials
• Press releases
• Open-source code in public repositories
• Public ISMS policies
• Public API documentation
• Blog posts
• Product brochures

MUST:

• Mark as "PUBLIC" in metadata/headers
• Allow unrestricted access
• Can be stored on public cloud services
• Can be transmitted unencrypted (though HTTPS is still recommended)
• Apply standard backup procedures
• Review before publication to ensure no unintended disclosure

MUST NOT:

• Include any Internal, Confidential, or Restricted information
• Include information that could aid attackers (detailed security configurations, vulnerability details)

INTERNAL Data

Definition: Information for use within Hack23 that should not be publicly disclosed but is not highly sensitive.

Examples:

• Internal documentation
• Meeting notes
• Project plans
• Employee directories
• Internal policies
• Development roadmaps
• Non-confidential customer communications
• Internal training materials

MUST:

• Mark as "INTERNAL" in metadata/headers
• Require authentication to access
• Store on approved internal systems or cloud services with access controls
• Transmit over encrypted channels (TLS 1.2+)
• Include in regular backup procedures
• Remove access when employees leave
• Apply retention policies

MUST NOT:

• Share with external parties without approval
• Store on personal devices without encryption
• Transmit over unencrypted channels
• Post on public websites or forums

CONFIDENTIAL Data

Definition: Sensitive information that could cause significant harm if disclosed.

Examples:

• Customer personal data (PII)
• Financial records
• Contracts and agreements
• Strategic plans
• Security assessment reports
• Non-public vulnerability information
• Source code for proprietary systems
• Audit reports
• Legal documents
• HR records (non-sensitive)

MUST:

• Mark as "CONFIDENTIAL" in metadata/headers
• Require strong authentication (MFA recommended)
• Implement role-based access control
• Encrypt at rest (AES-256)
• Encrypt in transit (TLS 1.2+)
• Log all access
• Store only on approved systems with encryption
• Apply strict retention and disposal policies
• Require NDA before sharing with external parties
• Conduct access reviews quarterly
• Remove access immediately when no longer needed

MUST NOT:

• Share via unsecured channels (email without encryption, chat without E2E encryption)
• Store on personal devices without full disk encryption and MDM
• Include in logs or error messages
• Share on public cloud services without encryption
• Copy to removable media without encryption and approval

RESTRICTED Data

Definition: Highly sensitive information that could cause severe harm if disclosed.

Examples:

• Passwords, encryption keys, API secrets
• Authentication credentials
• Personal health information (PHI)
• Financial account credentials
• Credit card data (PAN, CVV)
• National ID numbers (SSN, passport numbers)
• Biometric data
• Cryptographic private keys
• Security vulnerability details (before patching)

MUST:

• Mark as "RESTRICTED" in metadata/headers
• Require MFA for all access
• Implement strict role-based access control with separation of duties
• Encrypt at rest with AES-256
• Encrypt in transit with TLS 1.2+
• Log and monitor all access with alerts
• Store in dedicated secure systems (key management systems for secrets)
• Apply immediate retention (delete when no longer needed)
• Require explicit approval for each access
• Conduct access reviews monthly
• Use secret management systems (never hardcode)
• Implement automated rotation where possible
• Revoke access immediately when no longer needed

MUST NOT:

• Store in plaintext
• Include in source code, configuration files, or version control
• Include in logs, error messages, or debug output
• Transmit via email, chat, or unsecured channels
• Store on personal devices
• Copy to removable media
• Share screenshots or copies
• Display on screen in public areas
• Print unless absolutely necessary (and securely destroy after use)

Data Handling Matrix

Classification Guidelines

MUST:

• Classify all data at creation
• Apply the highest classification if multiple types are present
• Document classification decisions
• Review classification periodically (annually minimum)
• Train employees on classification
• Mark documents/files with classification labels
• Handle data according to its classification throughout its lifecycle

MUST NOT:

• Downgrade classification without approval
• Mix classifications in single document/database
• Ignore classification in automated systems

Storage Requirements

PUBLIC:

• Any approved storage

INTERNAL:

• Approved internal file shares
• Cloud storage with authentication
• Internal databases

CONFIDENTIAL:

• Encrypted cloud storage (approved providers)
• Encrypted databases
• Encrypted file systems
• Encrypted backups

RESTRICTED:

• Dedicated secure storage systems
• Key management systems (for secrets)
• HSM (for cryptographic keys)
• Encrypted backups in secure locations

Transmission Requirements

PUBLIC:

• Any method (HTTPS recommended)

INTERNAL:

• TLS 1.2+ for network transmission
• Authenticated email
• VPN for remote access

CONFIDENTIAL:

• TLS 1.2+ mandatory
• Encrypted email (S/MIME or PGP)
• Secure file transfer (SFTP, HTTPS)
• VPN for remote access

RESTRICTED:

• TLS 1.2+ mandatory
• End-to-end encrypted channels
• Out-of-band key exchange
• Never via standard email

Disposal Requirements

PUBLIC:

• Standard deletion

INTERNAL:

• Secure deletion (overwrite)
• Shred physical documents

CONFIDENTIAL:

• Secure deletion with verification
• Multiple overwrites
• Cross-cut shred physical documents
• Degauss magnetic media

RESTRICTED:

• Cryptographic erasure
• Physical destruction of media
• Certificate of destruction
• Immediate deletion from all locations

Examples

Example 1: Data Classification Headers (HTML)

<!-- PUBLIC data -->
<!DOCTYPE html>
<html lang="en">
<head>
  <meta name="classification" content="PUBLIC">
  <title>Hack23 - Cybersecurity Consulting</title>
</head>

<!-- INTERNAL data -->
<!DOCTYPE html>
<html lang="en">
<head>
  <meta name="classification" content="INTERNAL">
  <meta name="robots" content="noindex, nofollow">
  <title>Internal Project Documentation</title>
</head>

Example 2: Database Field Classification

// Define classification metadata
const userSchema = new mongoose.Schema({
  // PUBLIC
  username: {
    type: String,
    classification: 'PUBLIC',
    required: true
  },

  // INTERNAL
  email: {
    type: String,
    classification: 'INTERNAL',
    required: true
  },

  // CONFIDENTIAL
  firstName: {
    type: String,
    classification: 'CONFIDENTIAL',
    required: true
  },
  lastName: {
    type: String,
    classification: 'CONFIDENTIAL',
    required: true
  },
  dateOfBirth: {
    type: Date,
    classification: 'CONFIDENTIAL'
  },

  // RESTRICTED
  passwordHash: {
    type: String,
    classification: 'RESTRICTED',
    required: true,
    select: false // Never return in queries by default
  },
  ssn: {
    type: String,
    classification: 'RESTRICTED',
    select: false,
    // Encrypt at application level before storage
    get: (value) => decrypt(value),
    set: (value) => encrypt(value)
  }
});

Example 3: Logging with Classification

const winston = require('winston');

// Create logger with classification filter
const logger = winston.createLogger({
  format: winston.format.combine(
    winston.format.timestamp(),
    winston.format.json()
  ),
  transports: [
    new winston.transports.File({ filename: 'application.log' })
  ]
});

// Safe logging helper
function logSafely(level, message, metadata = {}) {
  // Filter out sensitive data
  const safeMetadata = {};

  for (const [key, value] of Object.entries(metadata)) {
    const classification = value?.classification || 'INTERNAL';

    // NEVER log RESTRICTED data
    if (classification === 'RESTRICTED') {
      safeMetadata[key] = '[RESTRICTED - REDACTED]';
    }
    // Mask CONFIDENTIAL data
    else if (classification === 'CONFIDENTIAL') {
      safeMetadata[key] = maskConfidential(value);
    }
    // INTERNAL and PUBLIC can be logged
    else {
      safeMetadata[key] = value;
    }
  }

  logger.log(level, message, safeMetadata);
}

function maskConfidential(value) {
  if (typeof value === 'string' && value.length > 4) {
    return value.substring(0, 2) + '***' + value.substring(value.length - 2);
  }
  return '***';
}

// Usage
logSafely('info', 'User login', {
  username: { value: 'john_doe', classification: 'PUBLIC' },
  email: { value: 'john@example.com', classification: 'INTERNAL' },
  ipAddress: { value: '192.168.1.1', classification: 'INTERNAL' },
  password: { value: 'secretpassword', classification: 'RESTRICTED' } // Will be redacted
});

// Output: { username: 'john_doe', email: 'john@example.com', ipAddress: '192.168.1.1', password: '[RESTRICTED - REDACTED]' }

Example 4: API Response Classification

// Filter API responses based on user permissions and classification
function filterResponseByClassification(data, userRole) {
  // Define role clearance levels
  const clearanceLevels = {
    'public': ['PUBLIC'],
    'user': ['PUBLIC', 'INTERNAL'],
    'admin': ['PUBLIC', 'INTERNAL', 'CONFIDENTIAL']
    // RESTRICTED data is never included in API responses
  };

  const allowedClassifications = clearanceLevels[userRole] || ['PUBLIC'];

  const filtered = {};
  for (const [key, value] of Object.entries(data)) {
    const classification = value?.classification || 'INTERNAL';

    if (classification === 'RESTRICTED') {
      // Never include RESTRICTED data in API responses
      continue;
    }

    if (allowedClassifications.includes(classification)) {
      filtered[key] = value.data || value;
    }
  }

  return filtered;
}

// Usage
app.get('/api/user/:id', requireAuth, async (req, res) => {
  const user = await db.users.findById(req.params.id);

  const userData = {
    username: { data: user.username, classification: 'PUBLIC' },
    email: { data: user.email, classification: 'INTERNAL' },
    firstName: { data: user.firstName, classification: 'CONFIDENTIAL' },
    lastName: { data: user.lastName, classification: 'CONFIDENTIAL' },
    ssn: { data: user.ssn, classification: 'RESTRICTED' } // Will never be included
  };

  const filtered = filterResponseByClassification(userData, req.user.role);
  res.json(filtered);
});

Example 5: File Classification Metadata

const fs = require('fs');
const xattr = require('fs-xattr');

// Set classification on file using extended attributes
async function setFileClassification(filePath, classification) {
  const validClassifications = ['PUBLIC', 'INTERNAL', 'CONFIDENTIAL', 'RESTRICTED'];

  if (!validClassifications.includes(classification)) {
    throw new Error(`Invalid classification: ${classification}`);
  }

  // Set extended attribute
  await xattr.set(filePath, 'user.classification', classification);

  // Set appropriate file permissions
  const permissions = {
    'PUBLIC': 0o644,      // rw-r--r--
    'INTERNAL': 0o640,    // rw-r-----
    'CONFIDENTIAL': 0o600, // rw-------
    'RESTRICTED': 0o600   // rw-------
  };

  await fs.promises.chmod(filePath, permissions[classification]);
}

// Get file classification
async function getFileClassification(filePath) {
  try {
    return await xattr.get(filePath, 'user.classification');
  } catch (err) {
    return 'INTERNAL'; // Default classification
  }
}

// Verify file handling based on classification
async function verifyFileHandling(filePath) {
  const classification = await getFileClassification(filePath);

  switch (classification) {
    case 'RESTRICTED':
    case 'CONFIDENTIAL':
      // Verify encryption
      const encrypted = await isFileEncrypted(filePath);
      if (!encrypted) {
        throw new Error(`${classification} file must be encrypted: ${filePath}`);
      }
      break;

    case 'INTERNAL':
      // Verify access controls
      const stats = await fs.promises.stat(filePath);
      if (stats.mode & 0o004) { // World readable
        throw new Error(`INTERNAL file must not be world-readable: ${filePath}`);
      }
      break;
  }
}

Example 6: Secure Deletion by Classification

const crypto = require('crypto');
const fs = require('fs').promises;

async function secureDelete(filePath, classification) {
  const stats = await fs.stat(filePath);
  const fileSize = stats.size;

  switch (classification) {
    case 'PUBLIC':
    case 'INTERNAL':
      // Simple deletion
      await fs.unlink(filePath);
      break;

    case 'CONFIDENTIAL':
      // Overwrite with random data 3 times
      for (let i = 0; i < 3; i++) {
        const randomData = crypto.randomBytes(fileSize);
        await fs.writeFile(filePath, randomData);
      }
      // Then delete
      await fs.unlink(filePath);
      break;

    case 'RESTRICTED':
      // Overwrite with random data 7 times (DoD 5220.22-M standard)
      for (let i = 0; i < 7; i++) {
        const randomData = crypto.randomBytes(fileSize);
        await fs.writeFile(filePath, randomData);
        // Sync to ensure write to disk
        const fd = await fs.open(filePath, 'r+');
        await fd.sync();
        await fd.close();
      }
      // Final overwrite with zeros
      await fs.writeFile(filePath, Buffer.alloc(fileSize, 0));
      // Then delete
      await fs.unlink(filePath);
      break;
  }

  // Log secure deletion
  logger.info('Secure deletion completed', {
    file: filePath,
    classification: classification,
    timestamp: new Date().toISOString()
  });
}

Related ISMS Policies

• Data Classification Policy - Primary policy for data classification
• Privacy Policy - Data protection and privacy requirements
• Cryptographic Controls Policy - Encryption requirements
• Access Control Policy - Access control by classification

Related Documentation

• CLASSIFICATION.md - Data classification guidance
• secure-development SKILL.md - Secure coding for classified data
• cryptography SKILL.md - Encryption for classified data

Compliance Mapping

ISO 27001:2022

• A.5.12 Classification of information
• A.5.13 Labelling of information
• A.5.14 Information transfer
• A.8.10 Information deletion

GDPR

• Article 32 - Security of processing (based on sensitivity)
• Article 17 - Right to erasure

NIST Cybersecurity Framework

• PR.DS-5: Protections against data leaks are implemented
• PR.IP-2: A System Development Life Cycle manages systems based on data sensitivity

Enforcement

All data MUST be classified. Violations:

• Unclassified data: Default to INTERNAL until properly classified
• Misclassification: Review and correct immediately
• Improper handling: Incident response, remediation required
• Unauthorized disclosure: Security incident, potential breach notification