🛡️ BLA (Blind LLM Agents) — Adversarial defense shield that protects binaries, Electron apps, JS bundles, Python, and WebAssembly against AI reverse-engineering agents like REA. Poison -> Exhaust -> Hallucinate.
JavaScript adversarial-aiclaude-codecursorelectron-securityllm-securitymcpmcp-serverobfuscationreverse-engineeringwebassembly
docs(architecture): add Red (morluto/rea) vs Blue (bla-agents) defense manifesto
qybaiheba0a9421c98f
diff --git a/docs/red-vs-blue-manifesto.md b/docs/red-vs-blue-manifesto.mdnew file mode 100644index 0000000..0148af3--- /dev/null+++ b/docs/red-vs-blue-manifesto.md@@ -0,0 +1,12 @@+# Red (`morluto/rea`) vs. Blue (`bla-agents`) — The Adversarial Defense Manifesto++- **REA (`re` = Red)**: Offensive reverse-engineering MCP (`Decompile -> Understand -> Recreate`).+- **BLA (`bl` = Blue / Blind LLM Agents)**: Defensive counter-reverse-engineering MCP (`Poison -> Exhaust -> Hallucinate`).++## Why Traditional Obfuscation Fails Against LLM Agents++Traditional obfuscators rename `calculatePrice` to `_0x1a2b`. Large language models trivially recover the original semantics by reading the surrounding control flow and constants.++BLA instead targets the **two structural bottlenecks of AI reverse-engineering agents**:+1. **Deterministic Parser & Context Budgets (Resource Exhaustion)**: Every static analyzer and MCP server in `morluto/rea` enforces strict recursion, candidate, and byte caps to prevent worker OOMs. BLA trips these exact caps in $O(1)$ runtime.+2. **Evidence-Driven Reasoning Discipline (Model Misjudgment Induction)**: Coding agents instructed to cite `observed` tool outputs over `unknown` paths will faithfully reconstruct BLA's crystal-clear **Shadow Business Logic** whenever the real business logic is guarded behind semantic `unknown` frontiers.