Back to Ruflo

Security SOTA Report — 2026-07-01

v3/docs/dream-cycle/2026-07-01-security-sota.md

3.34.05.7 KB
Original Source

Security SOTA Report — 2026-07-01

TL;DR: Six 2026 Grade A papers prove Ruflo's AgentDB retrieval pipeline is a high-confidence attack surface: certified defenses cut agent memory poisoning from 93–100% ASR to 0%, but Ruflo has neither a retrieval-layer injection filter nor behavioral forensics — and MCP caller identity remains unauthenticated across all major frameworks.

What's New in 2026

FindingSourceConfidence
SMSR certified defense cuts multi-session RAG poisoning from 93–100% → 0%arXiv:2606.12703 (Jun 2026)A
Forensic Trajectory Signatures detect memory poisoning: AUC=0.9904, generalizes across frontier models without retrainingarXiv:2606.30566 (Jun 2026)A
MIRROR: image poisoning ASR 76% vs 52% baseline; ships ART-SafeBench (41,000+ records)arXiv:2606.26793 (Jun 2026)A
Capability Gates Are Not Authorization: all 3 major frameworks (LangChain, LlamaIndex, Stripe Toolkit) conflate tool exposure with authorization; no fail-closed per-call gate by defaultarXiv:2606.28679 (Jun 2026)A
MCP Runtimes: naive baseline fails 10/10 attack cases; HCP mitigation blocks all 10arXiv:2606.29073 (Jun 2026)A
MCP Caller Identity Confusion: authorization state persists across callers with no re-auth gate; large-scale audit findingarXiv:2603.07473 (Mar 2026)B
Indirect Prompt Injection in the Wild: decomposes payloads into retrieval-guaranteed fragments; near-100% retrieval success across 11 benchmarksarXiv:2601.07072 (Jan 2026)A
OWASP LLM Top 10 2025: two new categories — LLM07 (System Prompt Leakage), LLM08 (Vector and Embedding Weaknesses)OWASP GenAI 2025B

Ruflo Current Capability

LayerCurrent StateGap
AgentDB write pathWrites accepted from any swarm agent without content verificationNo poison detection before commit
AgentDB retrieval pathHNSW retrieves top-K, injects directly into agent contextNo injection filter; OWASP LLM08 not addressed
MCP caller identityNo caller authentication; SafeExecutor validates format/syntax onlyAuthorization persists without re-auth (arXiv:2603.07473 pattern)
Memory anomaly detectionNoneNo behavioral forensics; AUC=0.9904 detection proven feasible
System prompt leakageNo guardrail on prompt exfiltration via tool outputsOWASP LLM07 not addressed

Competitor Comparison

FrameworkRAG/Memory Injection DefenseMCP Caller AuthBehavioral ForensicsLatest Version
LangGraphNo retrieval-layer filter; only output bleachingNo native caller authNo1.2.7 (Jun 2026)
AutoGen / AG2Docker isolation (code only); no RAG layer defenseNoNo0.12.x Beta
CrewAINo documented defenseNoNo1.14.2 (Apr 2026)
OpenAI Swarm (deprecated)N/A — ephemeral context onlyN/ANoOct 2024
Ruflo (current)None (ADR-165 closes CVE-1–5, not retrieval-layer)NoneNone3.16.2

All five frameworks share the RAG injection and caller-identity gaps. Ruflo is unique in having an extensible hook system (post-retrieve, pre-context-inject) that could be the insertion point — but the hooks do not exist yet.

Benchmarks

BenchmarkMetricGradeSource
SMSR certified defense — multi-session RAG agentASR: 93–100% (undefended) → 0% (SMSR, all unsigned variants)AarXiv:2606.12703, Jun 2026
Forensic Trajectory Signatures — cross-model hold-outAUC=0.9904 overall; AUC=1.000 on 6/9 frontier-model splitsAarXiv:2606.30566, Jun 2026
MIRROR multimodal RAG attackImage poisoning ASR 76% vs 52% baselineAarXiv:2606.26793, Jun 2026
Indirect injection retrieval (WILD)Near-100% retrieval success across 11 benchmarksAarXiv:2601.07072, Jan 2026
HCP MCP runtime mitigation10/10 attacks blocked (vs 0/10 naive baseline)AarXiv:2606.29073, Jun 2026

SOTA Proof & Witness

FieldValue
Session commit1a887969548b58e77f7853fedca922fc1cb5c6aa
Report SHA-256df009e8f56b505a827689066cf53ee1761b75360f21dc227427e939179837135
Witness stampaa9f66df228edde5ec6a529b6afb79cb490941d4aaf92395d0ff553b3b8f01a0

Verifier: printf '%s%s' 'df009e8f56b505a827689066cf53ee1761b75360f21dc227427e939179837135' '1a887969548b58e77f7853fedca922fc1cb5c6aa' | sha256sum → must yield aa9f66df228edde5ec6a529b6afb79cb490941d4aaf92395d0ff553b3b8f01a0

  1. Add AgentDbRetrievalGuard to the read path (v3/@claude-flow/memory/src/agentdb/retrieval-guard.ts): run each retrieved chunk through a lightweight injection detector (regex + semantic boundary check) before injecting into agent context. Target: reduce retrieval-layer ASR from ~100% (current) toward SMSR's 0% certified bound. Priority: High (ADR-180).

  2. Implement MemoryPoisonForensics background worker (13th worker slot, priority: critical): record behavioral trajectory signatures on each AgentDB write sequence; flag anomalies where trust-change score exceeds 2σ baseline. SOTA benchmark: AUC=0.9904 is achievable without retraining across frontier models (arXiv:2606.30566).

  3. Add MCP caller-identity binding to SafeExecutor: every MCP tool invocation must carry an invocation-bound capability token (AIP-style, arXiv:2603.24775); SafeExecutor rejects calls without a fresh per-invocation token. This closes the caller-identity persistence gap (arXiv:2603.07473) and satisfies OWASP LLM07 (system prompt leakage via tool channel).