BoxPwnr-Traces

BoxPwnr-Traces BoxPwnr의 추적 및 벤치마크 결과로, 여러 보안 플랫폼에 걸쳐 있습니다. 각 추적에는 전체 LLM 상호작용, 실행된 명령어, 마크다운 보고서 + 공격 그래프, 통계 및 사용된 구성이 포함됩니다. 리더보드를 탐색하고, 대화형 웹 뷰어에서 실행을 재생하며, AI 생성 보고서를 읽어보세요: 🔬 BoxPwnr Traces& Benchmarks...

DeepTrap

DeepTrap English | 中文 敵対的な実行コンテキスト下における OpenClaw エージェントのオープンワールドセキュリティ評価。 DeepTrap は、OpenClaw エージェントが悪意のある実行コンテキスト圧力（改ざんされたワークスペースファイル、注入されたスキル、誤解を招くツールメタデータ、安全でないコマンドパス、仕込まれた秘密情報、エンコードされたペイロード）に耐えながら、良性のユーザータスクを完了できるかを評価するセキュリティベンチマークです。 公開リリースには、6 つのコンテキスト脆弱性クラス × 7 つの運用シナリオファミリー として構成された 42...

Ajar

Ajar：衡量智能体防御中的开放权限 一个智能体安全基准会报告两个数字：攻击成功率和良性效用，而这两个数字都是从已发生的运行中读取的。两者都没有说明防御在那些未被运行的路径上准备允许什么。Ajar 直接对此发问：对于每个良性任务，它构建该任务不需要的候选工具调用，在智能体可能行动的每个时间点将每个候选调用呈现给防御，并对防御所允许的、按危害加权的比例进行评分。 它挂接到一个已经存在的基准上，并复用该基准已经携带的内容来进行自我评分——它的任务、工具模式、参考解决方案和目标状态——因此新的宿主只需要一个适配器，新的防御只需要一个包装器。...

Piggybacking on Perception: Stealthy Concurrent Audio Prompt Injections against Multimodal LLM Agents

Large Language Model LLM-driven multimodal agents are increasingly deployed to execute autonomous tasks via continuous audio interaction. While this paradigm enhances interaction naturalness, it introduces a critical yet under-explored attack surface, as audio inputs inevitably contain...

Open Security Benchmark: Towards Autonomous Enterprise Cyber Defense

Enterprises are moving toward autonomous cyber defense: agentic AI that builds situational awareness of an organization's security state and reasons from it to assessments, decisions, and actions. This rests on a holistic view of the enterprise's security state, the continuous, cross-vendor pictu...

ContraFix: Agentic Vulnerability Repair Via Differential Runtime Evidence and Skill Reuse

Large language model LLM agents are increasingly used for automated vulnerability repair AVR, where repository-level reasoning enables them to inspect context and produce source-code patches. However, recent empirical results show that these agents still struggle with real-world vulnerabilities...

ARGUS: Defending LLM Agents against Context-Aware Prompt Injection

The rise of Large Language Model LLM agents, augmented with tool use, skills, and external knowledge, has introduced new security risks. Among them, prompt injection attacks, where adversaries embed malicious instructions into the agent workflow, have emerged as the primary threat. However,...

Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks

LLM agents are evolving rapidly, powered by code execution, tools, and the recently introduced agent skills feature. Skills allow users to extend LLM applications with specialized third-party code, knowledge, and instructions. Although this can extend agent capabilities to new domains, it creates...

Benchmarking Large Language Models for Zero-Shot and Few-Shot Phishing URL Detection

The Uniform Resource Locator URL, introduced in a connectivity-first era to define access and locate resources, remains historically limited, lacking future-proof mechanisms for security, trust, or resilience against fraud and abuse, despite the introduction of reactive protections like HTTPS...

Can MLLMs Detect Phishing? A Comprehensive Security Benchmark Suite Focusing on Dynamic Threats and Multimodal Evaluation in Academic Environments

The rapid proliferation of Multimodal Large Language Models MLLMs has introduced unprecedented security challenges, particularly in phishing detection within academic environments. Academic institutions and researchers are high-value targets, facing dynamic, multilingual, and context-dependent...

MCPSecBench: a Systematic Security Benchmark and Playground for Testing Model Context Protocols

Large Language Models LLMs are increasingly integrated into real-world applications via the Model Context Protocol MCP, a universal, open standard for connecting AI agents with data sources and external tools. While MCP enhances the capabilities of LLM-based agents, it also introduces new securit...

RAS-Eval: a Comprehensive Benchmark for Security Evaluation of LLM Agents in Real-World Environments

The rapid deployment of Large language model LLM agents in critical domains like healthcare and finance necessitates robust security frameworks. To address the absence of standardized evaluation benchmarks for these agents in dynamic environments, we introduce RAS-Eval, a comprehensive security...

LLMs Caught in the Crossfire: Malware Requests and Jailbreak Challenges

The widespread adoption of Large Language Models LLMs has heightened concerns about their security, particularly their vulnerability to jailbreak attacks that leverage crafted prompts to generate malicious outputs. While prior research has been conducted on general security capabilities of LLMs,...

Configure a Proper Number of Concurrent Unauthenticated SSH Connections

Without knowing the password, an attacker can set up a large number of concurrent connections that have not been authenticated to consume system resources. The number of concurrent unauthenticated SSH connections is not configured in openEuler by default. You are advised to configure the upper...

Do Not Use X11 Forwarding

The X11 forwarding function of SSH allows the GUI program of the remote host to be executed on the local host. If the X11 forwarding function is enabled, the attack surface is expanded and other users on the X11 server may attack the local host. If the function is not required in the service...

Configure the dmesg Access Permission Properly

The permission to access dmesg information is restricted. Unprivileged users cannot view system information. This prevents any one from obtaining sensitive information and attacking the system. Only processes with the CAPSYSLOG capability are allowed to access kernel logs. In this way, the least...

Disable IP Forwarding

If a node does not function as a gateway server, disable the IP forwarding function. Otherwise, attackers can use the node as a router. In the container scenario, if network packets need to be forwarded through the host, IP forwarding is allowed. SPDX-FileCopyrightText: 2025 Greenbone AG Some tex...

Disable SysRq

SysRq enables users with physical access to access dangerous system-level commands in a computer. Therefore, it is advised to restrict the usage of the SysRq function. If SysRq is not disabled, you can use the keyboard to trigger SysRq. As a result, commands may be directly sent to the kernel,...

Configure Proper Policies for OUTPUT of nftables

There are two occasions in which a server sends outgoing packets: 1. The local host process proactively connects to an external server, for example, performing an HTTP access, or sending data to a log server. 2. The local host responds to the external access to the local services. If no policy is...

Ensure Kernel SMEP is Enabled

Supervisor Mode Execution Prevention SMEP can be enabled to prevent the user-space code execution on the kernel. If SMEP is not enabled, attackers can execute the user-space code through kernel-mode code redirection, which increases the attack surface and reduces system security. SMEP is enabled ...