cve-2025-32711

PowerShell-based detection and remediation toolkit for CVE-2025-32711 (EchoLeak), a critical zero-click AI command injection vulnerability in Microsoft 365 Copilot. Scans email patterns, audit logs, and security configurations to identify and mitigate prompt injection attacks.

• CVE ID: CVE-2025-32711
• Nickname: EchoLeak
• CVSS Score: 9.3 (Critical)
• Type: LLM Scope Violation / AI Command Injection
• Affected Product: Microsoft 365 Copilot
• Status: Patched server-side by Microsoft in May 2025
• User Action Required: None (server-side patch applied)

Attack Mechanism

The vulnerability exploits how Microsoft 365 Copilot processes and retrieves data:

1. Zero-Click Attack: Attacker sends a specially crafted email containing hidden prompt injections
2. Bypass Security: The email bypasses XPIA (Cross-Prompt Injection Attack) classifiers by appearing as user instructions
3. Data Exfiltration: When Copilot processes any query, it accesses the malicious email and executes the hidden instructions
4. Silent Leakage: Sensitive data from Copilot's context (emails, OneDrive, SharePoint, Teams) is sent to attacker-controlled servers

Repository Structure

Prerequisites

• Windows PowerShell 5.1 or PowerShell 7+
• Administrative privileges on the system
• Microsoft 365 PowerShell modules:
  • ExchangeOnlineManagement
  • Microsoft.Graph
  • AzureAD or AzureAD.Standard.Preview

Installation

1. Clone this repository:

2. Install required PowerShell modules:

3. Connect to Microsoft 365:

Usage

Detection Script

The detection script checks for:

• Microsoft 365 Copilot deployment status
• Suspicious email patterns indicating prompt injection attempts
• Security configuration vulnerabilities
• Audit log anomalies

Run the detection script:

Parameters:

• -OutputPath: Path for the detection report (default: .\logs\detection_report.json)
• -IncludeEmailAnalysis: Perform deep email analysis (may take longer)
• -Days: Number of days to look back in audit logs (default: 30)

Remediation Script

The remediation script implements:

• Enhanced email filtering rules
• DLP policies for prompt injection detection
• Audit logging improvements
• Security configuration hardening

Run the remediation script:

Parameters:

• -ConfigPath: Path to remediation configuration file
• -WhatIf: Preview changes without applying them
• -Force: Apply changes without confirmation prompts

Testing

Run the test suites to verify script functionality:

Detection Indicators

The tool looks for the following indicators of potential exploitation:

1. Email Patterns:

   • Emails containing markdown-style links with excessive query parameters
   • Messages with hidden instructions targeting AI assistants
   • Reference-style markdown links that bypass standard filters

2. Behavioral Anomalies:

   • Unusual data access patterns by Copilot service accounts
   • Spike in external data transfers
   • Suspicious audit log entries related to AI operations

3. Configuration Weaknesses:

   • Missing or misconfigured DLP policies
   • Inadequate email filtering rules
   • Disabled audit logging for AI operations

Remediation Actions

The remediation script performs the following actions:

1. Email Security:

   • Creates mail flow rules to detect prompt injection patterns
   • Implements content filtering for AI-specific threats
   • Blocks suspicious markdown patterns in emails

2. Data Loss Prevention:

   • Creates DLP policies for AI prompt injection detection
   • Monitors for data exfiltration patterns
   • Alerts on suspicious content combinations

3. Audit and Monitoring:

   • Enables comprehensive audit logging
   • Creates alert policies for AI-related anomalies
   • Configures retention policies for security logs

4. Access Controls:

   • Reviews and restricts Copilot data access permissions
   • Implements conditional access policies
   • Enforces principle of least privilege

Best Practices

1. Regular Monitoring: Run the detection script weekly to identify new threats
2. Update Patterns: Keep detection patterns updated as new attack variants emerge
3. Test Thoroughly: Always use -WhatIf parameter before applying remediation
4. Document Changes: Maintain logs of all security configurations applied
5. Stay Informed: Monitor Microsoft security advisories for updates

Limitations

• This tool cannot detect past exploitation if audit logs have been purged
• Server-side patches by Microsoft have already addressed the core vulnerability
• Detection patterns may need updates as attack techniques evolve
• Some remediation actions require specific Microsoft 365 licensing

Support

For issues, questions, or contributions:

• Open an issue in this repository
• Consult Microsoft's official documentation
• Review the Microsoft Security Response Center advisory

References

• Microsoft Security Advisory
• NVD Database Entry
• MITRE CVE Entry

License

This tool is provided as-is for security assessment purposes. Use at your own risk and ensure compliance with your organization's security policies.