HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems

HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems | Signal Canvas | ScienceToStartup

Page Freshness

Signal Canvas proof surface

Canonical route: /signal-canvas/hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems

degraded

Proof freshness: stale
Proof status: failed
Display score: 9/10
Last proof check: 2026-03-19
Score updated: 2026-04-02
Score fresh until: 2026-05-02
References: 0
Source count: 0
Coverage: 33%

This page has proof data, but the latest verification did not complete cleanly.

Agent Handoff

Canonical ID hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems | Route /signal-canvas/hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems

REST example

curl https://sciencetostartup.com/api/v1/agent-handoff/signal-canvas/hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems

MCP example

{
  "tool": "search_signal_canvas",
  "arguments": {
    "mode": "paper",
    "paper_ref": "hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems",
    "query_text": "Summarize HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems"
  }
}

source_context

{
  "surface": "signal_canvas",
  "mode": "paper",
  "query": "HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems",
  "normalized_query": "2602.22427",
  "route": "/signal-canvas/hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems",
  "paper_ref": "hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}

Evidence Receipt

Route status: degraded

Claims: 12

References: Pending verification

Proof: Verification pending

Freshness state: stale

Source paper: HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems

PDF: https://arxiv.org/pdf/2602.22427v1

Source count: Pending verification

Coverage: 33%

Last proof check: 2026-03-19T21:31:49.672Z

Signal Canvas receipt window

Watch and verify: HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems

/buildability/hubscan-detecting-hubness-poisoning-in-retrieval-augmented-generation-systems

Watchwatch

Subject: HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems

Verdict

Watch

Verdict is Watch because viability or proof quality is intermediate and should be re-evaluated before execution.

Preparing verified analysis

GitHub Code Pulse

No public code linked for this paper yet.

Claim map

Strong 12Mixed 0Weak 0

Evidencepartial
We introduce hubscan, an open-source security scanner that evaluates vector indices and embeddings to identify hubs in RAG systems.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
Hubscan presents a multi-detector architecture that integrates: (1) robust statistical hubness detection utilizing median/MAD-based z-scores, (2) cluster spread analysis to assess cross-cluster retrieval patterns, (3) stability testing under query perturbations, and (4) domain-aware and modality-aware detection for category-specific and cross-modal attacks.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
Our solution accommodates several vector databases (FAISS, Pinecone, Qdrant, Weaviate) and offers versatile retrieval techniques, including vector similarity, hybrid search, and lexical matching with reranking capabilities.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
hubscan achieves 90% recall at a 0.2% alert budget and 100% recall at 0.4%, with adversarial hubs ranking above the 99.8th percentile.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
Domain-scoped scanning recovers 100% of targeted attacks that evade global detection.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
Production validation on 1M real web documents from MS MARCO demonstrates significant score separation between clean documents and adversarial content.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
The system might encounter challenges with evolving adversarial tactics or new attack forms that bypass current detection methods. Continuous update and adaptation will be necessary.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
The need to secure RAG systems in enterprises using AI for decision support creates a large market valued in the cybersecurity sector, where companies will pay to ensure data integrity and system reliability.
Implicationmissing
Implication not extracted yet.
Verificationpartial
partial
Evidencepartial
Hubscan presents a multi-detector architecture that integrates: (1) robust statistical hubness detection utilizing median/MAD-based z-scores, (2) cluster spread analysis to assess cross-cluster retrieval patterns, (3) stability testing under query perturbations, and (4) domain-aware and modality-aware detection for category-specific and cross-modal attacks.
Implicationpartial
The abstract explicitly lists the components of the multi-detector architecture.
Verificationpartial
partial
Evidencepartial
hubscan achieves 90% recall at a 0.2% alert budget and 100% recall at 0.4%, with adversarial hubs ranking above the 99.8th percentile.
Implicationpartial
This is a specific quantitative result directly stated in the abstract.
Verificationpartial
partial
Evidencepartial
Our solution accommodates several vector databases (FAISS, Pinecone, Qdrant, Weaviate)
Implicationpartial
The abstract explicitly lists the supported vector databases.
Verificationpartial
partial
Evidencepartial
Domain-scoped scanning recovers 100% of targeted attacks that evade global detection.
Implicationpartial
This is a specific quantitative result directly stated in the abstract regarding a specific feature.
Verificationpartial
partial

Author intelligence and commercialization panels stay hidden until the proof receipt is verified, cites at least 3 references, includes at least 2 sources, and clears 50% coverage. The paper narrative and citation surfaces remain public while verification is pending.

HubScan: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems

Use Signal Canvas as the narrative proof surface

Use this Signal Canvas via API or MCP

Signal Canvas proof surface