moizxsec

// vulnerability research · appsec · offensive engineering

Abdul Moiz

Offensive security engineer. I find ways to break web applications, identity providers, cloud platforms, and the chains that hold them together. Then I help fix them.

Currently shipping AI-assisted application security at a HIPAA / SOC 2 healthcare platform. On the side, I disclose vulnerabilities in open-source software under coordinated disclosure.

Advisories
6
CVE IDs
6
Writeups
13
Critical / High
7 / 6

CVEs & advisories

6
view all

Vulnerabilities I found and reported to maintainers. Each row is a full writeup: root cause, proof of concept, fix, and the disclosure timeline.

CVE-2026-49950GHSA-mmvj-m246-m9vfXoscar: Unauthenticated RCE via Actor-Message Pickle DeserializationAffectedxoscar >= 0.1.0, <= 0.12.0PyPI · xorbitsai/xoscarfixed in 0.12.0 (opt-in authentication; not automatic)Severity Critical 9.3 WeaknessCWE-502Status Coordinated Date
CVE-2026-107709decompress-zip: Zip Slip Arbitrary File Write via a Sibling-Prefix BypassAffecteddecompress-zip ≤ 0.3.3 (no fix available)npm · bower/decompress-zipSeverity High 8.1 WeaknessCWE-22Status Coordinated Date
CVE-2026-59972GHSA-2756-j7rr-9464unzipper: Zip Slip Arbitrary File Write via a Sibling-Prefix Path BypassAffectedunzipper ≤ 0.12.4npm · ZJONSSON/node-unzipperfixed in 0.12.5Severity High 8.1 WeaknessCWE-22Status Patched Date
CVE-2026-55091GHSA-hp36-v28f-w3r4flat-to-nested: Prototype Pollution via a __proto__ Parent KeyAffectedflat-to-nested ≤ 1.1.1npm · joaonuno/flat-to-nested-jsfixed in 1.1.2Severity High 7.5 WeaknessCWE-1321Status Patched Date
CVE-2026-54283GHSA-82w8-qh3p-5jfqStarlette: Form-Parser Limits Silently Ignored for URL-Encoded BodiesAffectedstarlette >= 0.4.1, < 1.3.1PyPI · encode/starlettefixed in 1.3.1Severity High 7.5 WeaknessCWE-770Status Patched Date
CVE-2026-63358FileGator v7.14.0: Privilege Escalation via Unvalidated chmod EndpointAffectedFileGator < 7.14.2Packagist · filegator/filegatorfixed in 7.14.2Severity High 7.3 WeaknessCWE-732Status Patched Date

Open-source tooling

1
view all

Security tools I build and maintain in the open. Offensive by design, built to prove a weakness before someone else does, and to give defenders something to run first.

Author TypeScript MIT Published

mcpwn

An interception proxy, scanner, and guard for Model Context Protocol (MCP) traffic.

mcpwn sits on the wire between an AI agent and its MCP servers. It is a stdio and HTTP man-in-the-middle proxy, a self-driving scanner, and a defensive guard, built to find and demonstrate indirect prompt injection, tool poisoning, rug pulls, sampling abuse, and data exfiltration. Sometimes described as Burp Suite for AI agents.

  • stdio + HTTP MITM proxy
  • Live intercept and tamper
  • Agent-free active scanner
  • Guard firewall, fail-open
  • SARIF + CI gating
  • CVSS v3.1 + CWE advisories
  • TypeScript
  • Node >= 18
  • ESM
  • Zero runtime dependencies

Research & writeups

7
view all

Focus

about
Web · API
OWASP, NIST 800-115, PTES, OSSTMM
Identity
Active Directory, Entra ID, SAML, OAuth 2.0, OIDC, JWT
Cloud
AWS WAF, Azure, Firebase, Power Platform misconfiguration
AI / agents
Prompt-injection defence, dual-LLM cascades, automated CVE triage
Coordinated disclosure

90-day window from vendor acknowledgement. If you maintain something I've written about and want to talk, email is the most reliable channel.

muezzism@gmail.com