The following versions of MCP KQL Server are currently supported with security updates:

We take security vulnerabilities seriously. If you discover a security issue, please report it responsibly.

DO NOT create a public GitHub issue for security vulnerabilities.

Instead, please report security vulnerabilities by:

1. Email: Send details to arjuntrivedi42@yahoo.com with subject line: [SECURITY] MCP KQL Server Vulnerability Report

2. GitHub Security Advisory: Use GitHub's private vulnerability reporting

Please include the following information in your report:

• Description: Clear description of the vulnerability
• Impact: Potential impact and severity assessment
• Reproduction Steps: Detailed steps to reproduce the issue
• Affected Versions: Which versions are affected
• Suggested Fix: If you have a suggested fix, include it
• Your Contact: How we can reach you for follow-up

Subject: [SECURITY] SQL Injection in query parameter

Description:
The execute_kql_query function does not properly sanitize the 'query' parameter,
allowing potential KQL injection attacks.

Impact:
An attacker could execute arbitrary KQL queries, potentially accessing
unauthorized data or causing denial of service.

Severity: HIGH

Affected Versions: 2.0.0 - 2.1.0

Reproduction Steps:
1. Call execute_kql_query with query parameter containing malicious KQL
2. The query is executed without sanitization
3. Attacker gains access to unintended data

Suggested Fix:
Implement query parameterization or strict input validation.

Contact: security-researcher@example.com

When using MCP KQL Server, follow these security best practices:

# DO: Use Managed Identity in production
from azure.identity import DefaultAzureCredential
credential = DefaultAzureCredential()

# DON'T: Hardcode credentials
# credential = ClientSecretCredential(tenant_id, client_id, "hardcoded_secret")  # BAD!

# DO: Use environment variables for sensitive config
import os
cluster_url = os.environ.get("KUSTO_CLUSTER_URL")
database = os.environ.get("KUSTO_DATABASE")

# DON'T: Hardcode sensitive values
# cluster_url = "https://mycompany.kusto.windows.net"  # Avoid in code

# DO: Validate and sanitize user inputs
def safe_query(table_name: str, limit: int) -> str:
    # Validate table name against known tables
    if table_name not in allowed_tables:
        raise ValueError("Invalid table name")
    # Use parameterized limits
    if not isinstance(limit, int) or limit < 1 or limit > 10000:
        raise ValueError("Invalid limit")
    return f"{table_name} | take {limit}"

# DON'T: Directly interpolate user input
# query = f"{user_input} | take {user_limit}"  # Dangerous!

# DO: Sanitize logs
logger.info("Query executed for database: %s", database_name)

# DON'T: Log sensitive data
# logger.info(f"Query: {query}, Credentials: {credentials}")  # BAD!

MCP KQL Server includes these security features:

{
  "mcpServers": {
    "kql-server": {
      "command": "python",
      "args": ["-m", "mcp_kql_server"],
      "env": {
        "KQL_QUERY_TIMEOUT": "300",
        "KQL_MAX_ROWS": "10000",
        "KQL_LOG_LEVEL": "INFO"
      }
    }
  }
}

Note: This server uses Azure CLI authentication (az login). No service principal credentials (AZURE_CLIENT_ID, AZURE_TENANT_ID, AZURE_CLIENT_SECRET) are required. Simply ensure you are logged in via az login before starting the server.

1. Query Content: The server executes KQL queries as provided. Ensure queries are from trusted sources.

2. Local Cache: Schema information is cached locally in SQLite. Ensure appropriate file system permissions.

3. Log Output: Query errors may include partial query text in logs. Configure log levels appropriately.

Security updates are released as:

• Critical: Patch release within 7 days
• High: Patch release within 14 days
• Medium: Included in next minor release
• Low: Included in next major release

# Check for updates
pip index versions mcp-kql-server

# Update to latest
pip install --upgrade mcp-kql-server

# Enable auto-update checks (v2.1.0+)
# The server automatically checks for updates at startup

• No customer data is transmitted outside of Azure Data Explorer connections
• Schema metadata is cached locally for performance
• Query results are returned to the MCP client only
• No telemetry or analytics data is collected

For compliance requirements, enable detailed logging:

import logging
logging.getLogger("mcp_kql_server").setLevel(logging.DEBUG)

• Security Issues: arjuntrivedi42@yahoo.com
• General Questions: Open a GitHub issue
• Repository: https://github.com/4R9UN/mcp-kql-server

Thank you for helping keep MCP KQL Server secure!