PulseAugur
EN
LIVE 23:23:14
中文(ZH) 腾讯DSH安全论文实测：间接Prompt注入成功率17-25.5%，运行时防护怎么做？

Tencent paper reveals 25.5% indirect prompt injection success in DSH

By PulseAugur Editorial · [1 sources] ·

A recent security paper from Tencent's AI-Infra-Guard team highlights significant vulnerabilities in the DeepSeek Harness (DSH) open-source agent orchestration framework. The research found that indirect prompt injection attacks, where malicious instructions are embedded in external data sources, have a success rate of 17-25.5%. The paper also points out issues with unauthorized tool execution and a lack of validation for agent outputs and cross-tool data flows. To address these, the Correctover security plugin has been developed, offering runtime validation for agent frameworks like DSH. AI

IMPACT Highlights critical security gaps in agent orchestration frameworks, necessitating runtime validation for safe deployment.

RANK_REASON Security paper analyzing vulnerabilities in an AI agent framework. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — MCP tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Tencent paper reveals 25.5% indirect prompt injection success in DSH

How we ranked this

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Security paper analyzing vulnerabilities in an AI agent framework. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
safety, paper, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
49 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — MCP tag TIER_1 中文(ZH) · correctover ·

    Tencent DSH Security Paper Actual Test: Indirect Prompt Injection Success Rate 17-25.5%, How to Do Runtime Protection?

    <h1> 腾讯DSH安全论文实测：间接Prompt注入成功率17-25.5%，运行时防护怎么做？ </h1> <blockquote> <p>DeepSeek Harness（DSH）作为近期最火的开源Agent编排框架，GitHub星标破万。但腾讯AI-Infra-Guard团队刚发布的安全评估论文（arXiv:2608.16393）揭示了一个严峻事实：在受控测试中，间接Prompt注入攻击成功率高达17%-25.5%。</p> </blockquote> <h2> 一、论文核心发现 </h2> <p>腾讯团队对DSH进行了系统性安全评估，主要发现包括…

RELATED ENTITIES

RELATED TOPICS