一个存在22年的网络安全漏洞在短短两天内被AI检测出来，随后黑客攻击尝试大幅增加1。这个被称为PHP2Shell的漏洞预计影响全球70%以上的互联网网站2。Calif公司和Wiz公司分别利用AI前沿模型发现了两个PHP远程代码执行漏洞，其中一个PHP漏洞在2024年就已被发现但长期未被重视，直到大语言模型能够将理论漏洞转化为可工作的利用代码2。OpenAI的网络安全实验室研究人员正在参与这一危机的防护工作1。
过去一周内互联网安全形势严峻，多起重大漏洞事件接踵而至2。从安全补丁宣布到在野利用的时间已从过去的数周或数月级别大幅缩短至数小时或分钟级别2。同期发生的其他安全事件包括：GitLab分别在10月12日和21日发布两个CVSS评分为10.0的远程代码执行漏洞2；eMudhra证书颁发机构因使用极弱的私钥被Mozilla阻止8个中间证书2；LiteLLM遭到第二次恶意攻击，恶意包上传至PyPI2；PingCAP遭到未授权访问，网络勒索组织FulcrumSec声称泄露10.6TB数据2；Linux内核漏洞报告被误发至公开邮件列表而非私密列表，其中包含多个CVSS评分9.0以上的高危漏洞2。
A cybersecurity flaw that remained undetected for 22 years has been identified and rapidly weaponized, triggering a sharp escalation in hacking attempts across the internet.1 The vulnerability, known as PHP2Shell (CVE-2026-109618 and CVE-2026-109619), is projected to impact over 70 percent of websites globally.2 Researchers employed advanced artificial intelligence models to discover the flaw and convert theoretical exploits into functional attack code, compressing the timeline from security patch announcement to active exploitation from weeks or months down to hours or minutes.2
The acceleration of threat development underscores a broader pattern of security incidents that emerged over the past week.2 A PHP vulnerability originally identified in 2024 had been overlooked until large language models enabled researchers to transform it into a working exploit.2 Simultaneously, multiple other critical security breaches surfaced: GitLab released patches for two remote code execution vulnerabilities with CVSS scores of 10.0 on October 12 and 21; the eMudhra certificate authority was blocked by Mozilla after using extremely weak private keys such as "115," "56," and "4"; LiteLLM experienced a second compromise incident when malicious packages were uploaded to PyPI; and PingCAP suffered unauthorized access with the ransomware group FulcrumSec claiming to have exfiltrated 10.6 terabytes of data.2 Additionally, a critical Linux kernel vulnerability with a CVSS score exceeding 9.0 was accidentally disclosed to a public mailing list rather than being handled through private channels.2 Cybersecurity researchers from organizations including OpenAI's Cyber Lab are racing to implement protective measures against these threats.1
评论
还没有评论，欢迎留下第一条。