Matthew Keeley is a distinguished security engineer and researcher with over a decade of experience in platform security, AI-driven vulnerability research, and security automation. Matthew has been instrumental in building secure-by-default platform architectures and discovering critical vulnerabilities that have shaped industry security practices.
Matthew founded Platform Security to share practical field experience and cutting-edge research with the security community. His work has led to the discovery of over 20 CVEs, development of automated vulnerability detection systems, and pioneering research in adversarial AI testing. He regularly contributes to open source security tools and shares methodology through detailed technical write-ups.
# Posts by Matthew
React2Shell for Lambdas
January 26, 2026
Existing scanners miss CVE-2025-55182 in serverless Lambda deployments. While traditional RCE is blocked by Webpack bundling, the vulnerability enables Server Side JavaScript Injection (SSJI) that can exfiltrate AWS credentials, often more dangerous than shell access in cloud environments.
The GCP Service Account That Could Read Everything
December 20, 2025
A short story about Artifact Registry reader access, container images, and why build history quietly leaks secrets.
Hawk: Automated Credential Harvesting in WRCCDC
November 23, 2025
How a tiny ptrace bird turned into a credential avalanche. The story of building Hawk, a lightweight Golang tool that silently intercepts SSH and sudo credentials in real time, and why it dominates in competitive red team exercises.
The Anatomy of a Phishing Landing Page
September 27, 2025
A research-driven look into how attackers construct phishing landing pages that mimic Microsoft federated login portals, and what defenders can do about it.
Redshift Authentication is a UX Dumpster Fire
July 13, 2025
Redshift authentication sounds simple until you try to make it secure and usable at scale. This post dives into the broken state of IAM, in-database users, and the messy tradeoffs data and platform teams face daily.
How I Used AI to Create a Working Exploit for CVE-2025-32433 Before Public PoCs Existed
April 17, 2025
A step-by-step walkthrough of how I leveraged AI to analyze, understand, and exploit the Erlang SSH pre-authentication vulnerability (CVE-2025-32433) without any existing public proof of concept. Learn how AI is transforming vulnerability research and exploit development.
Machine Learning Evasion Attacks: How Adversaries Trick AI Models
March 17, 2025
A deep dive into how adversarial attacks manipulate machine learning models, the types of evasion attacks, and real-world implications.