LN: Hou et al. (2025) — Model Context Protocol (MCP): Landscape, Security Threats, and Future Directions

Bibliographic Reference

Citation: Hou, X., Zhao, Y., Wang, S., & Wang, H. (2025). Model context protocol (MCP): Landscape, security threats, and future research directions. arXiv:2503.23278. https://arxiv.org/abs/2503.23278


Pass 1 — Bird’s Eye View (5 Cs)

CAssessment
CategorySurvey + security analysis
ContextAcademic analysis of the MCP standard introduced by Anthropic (Nov 2024)
CorrectnessBased on official MCP specification + empirical security analysis
Contributions(1) First academic survey of MCP landscape; (2) Security threat taxonomy (injection attacks, tool poisoning, data exfiltration); (3) Research directions for secure MCP deployments
ClarityGood. Accessible for non-security readers.

Relevance: ⭐⭐⭐

PUMA Stage 5 Smart PMO uses MCP for tool integration (Jira API, GitHub). Understanding MCP security is important for the governance layer design.


PUMA Connection

PUMA’s governance layer (OPA policies, bounded autonomy — Constitution Article 4) must address MCP security threats. This paper provides the threat taxonomy for the ethics review (Section 1.3) and architecture spec.

MOCs