<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Rag-Security on CuraSec</title><link>https://curasec.metacog.co.kr/tags/rag-security/</link><description>Recent content in Rag-Security on CuraSec</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 02 Nov 2026 21:42:19 +0000</lastBuildDate><atom:link href="https://curasec.metacog.co.kr/tags/rag-security/index.xml" rel="self" type="application/rss+xml"/><item><title>RAG Knowledge Poisoning Defenses Overstate Effectiveness via ASR Metric</title><link>https://curasec.metacog.co.kr/insights/2026-11-02-prevented-is-not-protected-what-attack-success-rate-hides-ab/</link><pubDate>Mon, 02 Nov 2026 21:42:19 +0000</pubDate><guid>https://curasec.metacog.co.kr/insights/2026-11-02-prevented-is-not-protected-what-attack-success-rate-hides-ab/</guid><description>&lt;ul>
&lt;li>&lt;strong>Engineer — Learn:&lt;/strong> If you&amp;rsquo;re deploying RAG pipelines, this research shows that defenses evaluated purely on attack success rate can still introduce unexpected target-positive outputs — worth reviewing your evaluation framework before trusting any poisoning defense you&amp;rsquo;ve implemented.&lt;/li>
&lt;li>&lt;strong>SOC/IR — Skip&lt;/strong>&lt;/li>
&lt;li>&lt;strong>Leader — Learn:&lt;/strong> Useful context for AI governance discussions: defenses on enterprise RAG systems may have overstated efficacy, which matters when briefing on AI risk posture or assessing vendor claims about RAG security controls.&lt;/li>
&lt;/ul></description></item></channel></rss>