Papers, a standards draft and technical posts, with the primary link for each.
2026-08-26Preprint, Zenodo
Never Emitted: Reporter Attribution in GitHub's Machine-Readable Vulnerability Records
A measurement of whether the name of the person who reported a vulnerability survives from the authoritative CVE record into GitHub's OSV export and the NVD. In the author's comparison of CVEs published 1 to 20 August 2026, every record with a credits field upstream had lost it in the NVD copy. A replication package with the four datasets is public.
Replication package: doi.org/10.5281/zenodo.22119153
2026-05-24Preprint, Zenodo
Protocol Pivoting: Cross-Protocol Attack Escalation in Agentic AI Systems
Agentic systems run several agent protocols in parallel: MCP for tool access, A2A for delegation between agents, and others. Each assumes it operates alone. The paper formalizes Protocol Pivoting, an attack where an adversary enters through one protocol, exploits trust assumptions between protocols, and escalates to capabilities reachable only through another. It presents three scenarios and proposes a cross-protocol trust boundary model with three protocol-agnostic mitigations.
2026-06-02IETF Internet-Draft, revision 00
Security Considerations for Model Context Protocol (MCP) Implementations in AI Agent Systems
draft-mohiuddin-mcp-security-considerations-00. A self-published Internet-Draft: it is not a standard, has not been adopted, and expires 2026-12-04 unless renewed.
2026-09-29dev.to
Four vendors, one bad assumption: SSRF in MCP servers
2026-05-23dev.to
ai-fix: when a command fails, one word fixes it
2026-05-20dev.to
The security problem nobody is talking about: MCP servers