Cybersecurity & AI Research
Technical research notes, threat breakdowns, and architectural evaluations published through Cyravex Labs and Cyravex Security.
Indirect Prompt Injection in RAG Architectures: Attack Vectors & Context Isolation
A technical evaluation of how untrusted context retrieved into LLM prompts can lead to unauthorized tool execution, with proposed isolation defenses.
Bypassing BOLA/IDOR in GraphQL & REST Microservices
Analysis of broken object-level authorization vulnerabilities missed by automated scanners and manual verification methodologies.
Engineering a Pragmatic 90-Day Security Program Baseline
Architectural and governance framework for high-growth tech companies preparing for enterprise customer vendor reviews.
Applied Research Focus Areas & Peer Standards
Our research focuses on high-impact threat surfaces where traditional signature-based detection and automated scanners fail.
Adversarial AI & Context Boundaries
Evaluating non-deterministic failure modes in LLM agent pipelines, testing vector retrieval boundaries against poisoned embedding attacks, and designing cryptographic context verification.
Stateful API Authorization (BOLA)
Investigating horizontal and vertical privilege escalation in complex GraphQL schemas, microservice service meshes, and distributed tenancy layers where state ownership logic diverges.
Coordinated Vulnerability Disclosure
Adhering to ethical 90-day coordinated disclosure standards when examining third-party frameworks, assisting open-source maintainers with patches prior to public advisory release.
Collaborate with Cyravex Research
Engage directly with our research fellows on vulnerability disclosure, adversarial AI safety, or cryptography benchmarks.