AI渗透测试

Caderno público · Scout

重新定义了AI渗透测试的概念，从资源利用转向行为目标违反，具有创新性。

itens

vanguarda

interessante

data

07-16

vanguarda · score 9

AI渗透测试

重新定义了AI渗透测试的概念，从资源利用转向行为目标违反，具有创新性。

source: Rethinking Penetration Testing for AI-Enabled Systems: From Resource Compromise to Behavioral Objective Violation

original source
https://arxiv.org/abs/2607.14006v1

vanguarda · score 9

SPyCE: 技能-策略共进化

一种新的强化学习方法，促进多模态代理的视觉证据迭代使用。

source: SPyCE: Skill-Policy Co-evolution for Multimodal Agents

original source
https://arxiv.org/abs/2607.13854v1

vanguarda · score 9

Hack em DeFi: Oracle ataca Ostium

Detalhes técnicos sobre uma exploração recente em DeFi, com vulnerabilidade de oracle.

source: Another DeFi Exploit: Perp DEX Ostium Loses $18 Million in Oracle Attack

original source
https://decrypt.co/373566/defi-exploit-ostium-oracle-hack

vanguarda · score 8

Segurança de Skills em Agentes LLM

Este trabalho aborda a segurança de skills reutilizáveis em agentes LLM, uma área emergente e pouco explorada.

source: Agent Skill Security: Threat Models, Attacks, Defenses, and Evaluation

original source
https://arxiv.org/abs/2607.13987v1

vanguarda · score 8

DeepStress: 深度压力测试

评估深度搜索代理在面对低质量证据时的鲁棒性，具有重要实际应用价值。

source: DeepStress: Stress-Testing Deep Search Agents

original source
https://arxiv.org/abs/2607.13920v1

interessante · score 7

Framework Distribuído para Contagem de Modelos SAT

Um novo framework para contagem de modelos SAT, melhorando a eficiência de sistemas de workflow.

source: Assessing the Forensic Viability of Android Memory Analysis Across Production Builds: A Cross-Version Study of Security Hardening and Structure Preservation

original source
https://arxiv.org/abs/2607.13821v1

interessante · score 7

Modelo de Telemetria para Riscos de Atualização

Um modelo inovador para quantificar riscos de atualização em workflows duráveis, melhorando a segurança.

source: A Telemetry-Driven Model for Quantifying Upgrade Risk in Durable Workflow Execution

original source
https://arxiv.org/abs/2607.13617v1

interessante · score 7

Impactos ambientais de IA autônoma

Estudo sobre os custos ambientais da implementação de IA autônoma em países em desenvolvimento.

source: The Environmental Cost of Digital Sovereignty: Water, Energy, and Emissions Impacts of Sovereign AI Infrastructure in the Global South

original source
https://arxiv.org/abs/2607.13443v1

interessante · score 7

Generative art flexibility

User discusses pushing generative art to be more aggressive, seeking feedback.

source: Not everything needs to be tidy and regular... I pushed this to be more aggressive. Comments welcome!

original source
https://www.reddit.com/r/generative/comments/1uxd1rz/not_everything_needs_to_be_tidy_and_regular_i/

interessante · score 7

Mandala generative art

User shares mandala art, a form of generative art, with potential for technical and artistic interest.

source: Mandala Art

original source
https://www.reddit.com/r/generative/comments/1uxuw91/mandala_art/

English edition. Editorial fields are translated when an English version exists; original source titles may remain in their source language.