The UnboundCompute blog
Writing on web security
-
SAML Signature Wrapping Explained: When a Valid Signature Lies
SAML signature wrapping lets a valid signature bless one element while your app reads another. Learn how this XML attack bypasses login…
-
Dependency Confusion Attack Explained
A dependency confusion attack tricks your package manager into pulling an attacker’s public package over your private one. Learn how to stop…
-
RAG Data Poisoning: How Attackers Corrupt the Knowledge Base Behind an LLM
RAG data poisoning lets attackers plant false facts or hidden instructions in a knowledge base so an LLM retrieves and trusts them.…
-
The lethal trifecta in AI agents
The lethal trifecta is when an AI agent holds private data, untrusted content, and an outbound channel at once. Break one leg…
-
JWT Algorithm Confusion Attack Explained
A JWT algorithm confusion attack tricks a server into verifying forged tokens. Learn the alg:none bug, the RS256 to HS256 key swap,…
-
AI in Security Testing: What It Actually Does and Where It Falls Down
A vendor neutral guide to ai in security testing: where AI genuinely helps with recon, fuzzing, logic flaws, and triage, and where…
-
LLM Security Testing Tools: A Vendor Neutral Landscape Guide
A vendor neutral map of llm security testing tools: tools that use LLMs to test, and tools that red team LLM apps,…
-
AI Security Testing: A Vendor Neutral Guide to Where AI Helps and Where It Fails
AI security testing uses AI to find and prove software vulnerabilities. See where it helps, where it fails, the tool categories, and…
-
The GraphQL Attack Surface: Introspection, Query DoS, Broken Authorization, and Injection
The graphql attack surface comes from letting the client shape the query. See how introspection, nested queries, and broken field authorization get…
-
What Is a Padding Oracle Attack and How It Decrypts CBC Without the Key
A padding oracle attack decrypts CBC ciphertext without the key by abusing one bit of padding feedback. See the byte at a…