Skip to content

Latest commit

 

History

History
104 lines (92 loc) · 5.27 KB

attack-tree.md

File metadata and controls

104 lines (92 loc) · 5.27 KB

Attack Tree Analysis for milvus-io/milvus

Objective: Exfiltrate sensitive vector data and/or disrupt the availability of the Milvus-backed application. [CN]

Attack Tree Visualization

                                     +-----------------------------------------------------+
                                     |  Attacker Goal: Exfiltrate Data OR Disrupt Service  | [CN]
                                     +-----------------------------------------------------+
                                                      /                 |
                                                     /                  |
          +--------------------------------+  +---------------------+
          |      Exfiltrate Data        |  |   Disrupt Service   |
          +--------------------------------+  +---------------------+
                 /       |                        /       |
                /        |                       /        |
+-------------+[HR] +-------+-----+ +-----+ +------+-----+
| Unauthorized|     | Exploit |     | DDoS| | Resource|
|  Access to |     |  Known  |     | [HR]| | Exhaustion|
|   Data     |     |  Milvus |     |     | |   [HR]   |
|    [CN]    |     |   CVEs  |     |     | |          |
+-------------+     +-------+-----+ +-----+ +------+-----+
      |               |                       |
      |[HR]           |[HR]                   |[HR]
+-----+-----+ +-------+-----+         +-----+-----+
|  Weak   | |  Default  |         |  High|
|  Creds  | |  Creds   |         |  Query|
|  [HR]   | |  [HR]   |         |  Load|
|         | |          |         | [HR] |
+-----+-----+ +-------+-----+         +-----+-----+

Attack Tree Path: Exfiltrate Data

  • Critical Node: Unauthorized Access to Data [CN]

    • Description: The attacker gains direct access to the vector data stored in Milvus without proper authorization. This is the primary target for data exfiltration.
    • Impact: Very High - Loss of sensitive data, potential regulatory violations, reputational damage.

  • High-Risk Path: Unauthorized Access to Data -> Weak Credentials [HR]

    • Description: The attacker uses weak, easily guessable, or compromised passwords to gain access to Milvus accounts with data access privileges.
    • Likelihood: Medium
    • Impact: High
    • Effort: Low
    • Skill Level: Low
    • Detection Difficulty: Medium
    • Mitigation:
      • Enforce strong password policies (length, complexity, regular changes).
      • Implement multi-factor authentication (MFA).
      • Monitor login attempts for suspicious activity (e.g., multiple failed logins).

  • High-Risk Path: Unauthorized Access to Data -> Default Credentials [HR]

    • Description: The attacker uses default credentials that were not changed after the initial Milvus installation.
    • Likelihood: Low (assuming basic security practices are followed)
    • Impact: High
    • Effort: Very Low
    • Skill Level: Very Low
    • Detection Difficulty: Low
    • Mitigation:
      • Mandatory change of default credentials upon installation.
      • Automated checks for default credentials during deployment.

  • High-Risk Path: Unauthorized Access to Data -> Exploit Known Milvus CVEs -> Unpatched Milvus Version [HR]

    • Description: The attacker exploits a publicly known vulnerability (CVE) in an unpatched version of Milvus to gain unauthorized access to data.
    • Likelihood: Medium (depends on patch management practices)
    • Impact: High
    • Effort: Medium
    • Skill Level: Medium
    • Detection Difficulty: Medium
    • Mitigation:
      • Establish a robust vulnerability management program.
      • Regularly update Milvus to the latest stable version.
      • Subscribe to Milvus security advisories.
      • Use vulnerability scanners to identify unpatched systems.

Attack Tree Path: Disrupt Service

  • High-Risk Path: Disrupt Service -> DDoS [HR]

    • Description: The attacker overwhelms the Milvus service with a large volume of requests, making it unavailable to legitimate users.
    • Likelihood: High
    • Impact: Medium
    • Effort: Low
    • Skill Level: Low
    • Detection Difficulty: Low
    • Mitigation:
      • Implement rate limiting.
      • Use a Web Application Firewall (WAF) with DDoS protection capabilities.
      • Deploy Milvus behind a load balancer.
      • Consider using a Content Delivery Network (CDN) to distribute traffic.

  • High-Risk Path: Disrupt Service -> Resource Exhaustion -> High Query Load [HR]

    • Description: The attacker sends a large number of complex or resource-intensive queries to Milvus, consuming excessive CPU, memory, or disk resources, leading to service degradation or unavailability.
    • Likelihood: Medium
    • Impact: Medium
    • Effort: Medium
    • Skill Level: Medium
    • Detection Difficulty: Medium
    • Mitigation:
      • Implement query quotas and resource limits.
      • Monitor resource usage and set up alerts for unusual activity.
      • Optimize queries for performance.
      • Implement query timeouts.
      • Consider using a query analysis tool to identify and block malicious queries.