内容来源：https://www.wired.com/story/security-news-this-week-hackers-are-posting-the-claude-code-leak-with-bonus-malware/

内容总结：

本周网络安全与地缘政治风险事件频发，引发全球关注。据美国国土安全部记录披露，去年秋季在芝加哥执行“中途岛闪电行动”的部分边境巡逻队准军事人员被指多次对平民使用武力，其中数名人员亦出现在其他州的类似行动中。同时，美国海关和边境保护局（CBP）因内部敏感设施门禁密码通过在线学习平台Quizlet意外泄露，暴露出安全管理漏洞。

科技领域方面，苹果公司罕见为旧版操作系统iOS 18发布安全补丁，以应对已在野外利用的DarkSword黑客技术。该技术可使iPhone在访问恶意网页时被植入攻击工具。此外，人工智能公司Anthropic旗下编程工具Claude Code源代码遭意外泄露，部分流传版本被黑客植入信息窃取恶意软件，该公司正通过版权投诉要求下架相关代码库。

地缘冲突持续冲击全球安全格局。美国与以色列同伊朗的冲突进入第二个月，伊朗威胁将对包括苹果、谷歌、微软在内的十余家美企在 Gulf 地区的设施发动攻击，霍尔木兹海峡航运受阻已对全球经济造成严重影响。另一方面，美国联邦调查局（FBI）首次将对其监控系统的网络入侵定性为“重大事件”，指其可能危及国家安全。美媒称中国疑似幕后主导，若查实将标志FBI重大反情报失误。该局近年已多次遭外国黑客侵入调查系统与监控基础设施。

网络安全事件方面，美国执法部门近日捣毁四个相互关联的大型僵尸网络，其中Aisuru与Kimwolf网络曾发动史上最大规模分布式拒绝服务攻击。22岁大学生本杰明·布伦戴奇通过追踪住宅代理设备漏洞，为破案提供关键线索。加密货币行业亦受重创：去中心化金融平台Drift遭黑客窃取2.8亿美元，区块链分析公司Elliptic指出来自朝鲜的黑客嫌疑最大，其今年累计盗窃金额已近3亿美元。此外，思科公司成为软件供应链攻击最新受害者，黑客通过漏洞扫描软件Trivy窃取其源代码及客户数据，TeamPCP黑客组织被指系幕后黑手。

（注：本报道基于公开信息整合，涉及地缘政治与安全指控仅为外媒表述，不代表本方立场。）

中文翻译：

《连线》杂志本周根据美国国土安全部记录展开的一项调查，揭露了去年秋天在芝加哥"中途岛突袭行动"中频繁对平民使用武力的准军事化边境巡逻队成员身份。《连线》发现其中多名特工曾出现在美国其他州的类似行动中。

美国海关与边境保护局或许该注意保护其敏感设施信息了。《连线》通过基础谷歌搜索，在在线学习平台Quizlet上发现用户制作的记忆卡片，其中包含海关与边境保护局设施的门禁密码等更多信息。

苹果公司本周罕见地为iOS 18系统发布了"向后移植"补丁，以保护数百万仍在使用旧操作系统的用户免受DarkSword黑客技术的侵害。该技术于三月被发现，能让攻击者通过访问加载了嵌入接管工具的网站来感染iPhone。苹果最初敦促用户升级至当前操作系统iOS 26，但在DarkSword持续扩散后最终发布了iOS 18补丁。

美以与伊朗的战争本周进入第二个月，伊朗威胁将对十余家美国企业发动攻击，包括在海湾地区设有办事处和数据中心的苹果、谷歌、微软等科技巨头。这场看不到尽头的致命冲突持续对全球经济造成严重破坏，航运队伍仍滞留在关键贸易通道霍尔木兹海峡。与此同时，有人开始担忧若美国袭击对伊朗核设施造成实质损害可能引发的后果。

这还不是全部！每周我们都会汇总未深入报道的安全与隐私新闻。点击标题阅读完整报道，请注意安全防护。

黑客正用Claude代码泄露夹带恶意软件
本周早些时候，安全研究人员指出Anthropic意外公开了其热门编程工具Claude Code的源代码。随即有人在开发者平台GitHub上转发了这些代码。但若想自行下载部分代码库请警惕：BleepingComputer报道称部分发布者实为黑客，他们在代码行中隐藏了信息窃取恶意软件。

Anthropic方面正通过发布版权删除通知来清除泄露副本（无论是否含恶意软件）。《华尔街日报》报道称该公司最初试图删除GitHub上8000多个代码库，后缩减至96个副本及改编版本。

这已非黑客首次利用人们对Claude Code的关注进行攻击。该工具要求可能不熟悉计算机终端的用户从网站复制粘贴安装命令。今年三月，404 Media曾报道谷歌赞助广告将用户引向伪装成官方Claude Code安装指南的网站，这些网站引导用户运行实际会下载恶意软件的命令。

FBI窃听工具遭黑客入侵被正式列为国家安全风险
美国联邦调查局正式将近期对其监控采集系统的网络入侵归类为《联邦信息安全现代化法案》规定的"重大事件"——该法律定性专用于可能对国家安全构成严重风险的违规行为。本周初向国会报告的这项认定，据信是该局至少自2020年以来首次宣布自身系统遭遇重大事件。Politico援引两名未具名的特朗普政府高级官员报道称，中国被认为是此次入侵的幕后黑手。若得到证实，此次入侵可能标志着FBI的重大反情报失误。

FBI表示二月在其网络上检测到"可疑活动"。在3月4日提交国会的通知中（该文件经Politico查阅），该局称受入侵系统为非保密级别，存有"法律程序反馈"，例如根据法庭命令收集的电话和互联网元数据，以及"与FBI调查对象相关的个人信息"。据报道入侵者通过商业互联网服务提供商获得访问权限，FBI称这种方法反映出"复杂策略"。在其唯一公开声明中，该局表示已部署"所有技术能力进行应对"。

此次入侵事件进一步印证了黑客（绝大多数若非全部则为外国势力）渗透FBI自身系统及监控基础设施的模式。2023年，一名外国黑客通过暴露的法医实验室服务器访问了该局爱泼斯坦调查案文件。上月，与伊朗有关的黑客入侵了FBI局长卡什·帕特尔个人邮箱。2024年曝光的"盐台风"行动中，中国黑客入侵至少八家国内电信和互联网服务提供商——利用了据信与当前入侵事件相关的同一监控基础设施的运营商端。FBI去年承认盐台风行动已入侵至少80个国家的200家公司，研究人员称其未见放缓迹象。

22岁大学生如何协助摧毁破纪录僵尸网络
两周前，美国执法部门宣布对四个相互关联的僵尸网络（被恶意软件劫持以执行黑客指令的大规模计算机集群）进行了里程碑式打击，这些网络分别名为Aisuru、Kimwolf、JackSkid和Mossad。特别是Aisuru和Kimwolf僵尸网络曾发动史上最大规模的分布式拒绝服务网络攻击，利用成群的被黑物联网设备向受害者发送垃圾流量。

《华尔街日报》近日详细报道了这些僵尸网络调查中一位出人意料的参与者——22岁的罗切斯特理工学院学生本杰明·布伦戴奇。布伦戴奇执着追踪Kimwolf僵尸网络，发现其通过充当"住宅代理"的设备感染了全球家庭网络，实质上为这些网络提供了后门。他甚至潜伏在Discord平台上与疑似掌握黑客行动内幕信息的人交谈，获取关键技术线索并分享给执法部门。《华尔街日报》在报道布伦戴奇故事的同时，还提供了实用指南帮助判断家庭网络是否易受住宅代理设备攻击及如何加强防护。

朝鲜黑客可能从Drift加密平台窃取2.8亿美元
鉴于近年来加密货币行业的不安全性持续为金正恩政权提供资金，2026年发生朝鲜大规模加密盗窃案已在预料之中。目前去中心化金融平台Drift承认遭遇网络安全漏洞，2.8亿美元被盗。加密追踪公司Elliptic根据黑客与被盗加密货币区块链的交互线索、"洗钱手法及网络层级指标"，将入侵事件归咎于朝鲜黑客。Elliptic表示朝鲜黑客今年总计窃取近3亿美元，其中绝大部分来自此次最新盗窃案。尽管数额巨大，该国黑客今年仍难以超过去年总计20亿美元的加密盗窃总额。

思科源代码在软件供应链攻击潮中被盗
网络安全媒体Bleeping Computer本周报道，思科成为软件供应链黑客攻击潮的最新受害者，导致该公司及部分客户的源代码被盗。此次入侵似乎是TeamPCP黑客组织所为，该组织已用自身恶意代码入侵多款安全软件，进而利用恶意软件获取的访问权限窃取用户凭证。据报道，此次思科凭证是通过漏洞扫描软件Trivy被入侵而失窃，使黑客得以访问思科开发环境。思科入侵事件只是TeamPCP为传播其信息窃取恶意软件而实施的一系列供应链攻击中的最新案例，此前还包括通过LiteLLM人工智能软件和安全软件CheckMarx进行的攻击。

英文来源：

A WIRED investigation based on Department of Homeland Security records this week revealed the identities of paramilitary Border Patrol agents who frequently used force against civilians during Operation Midway Blitz in Chicago last fall. Several of the agents, WIRED found, appeared in similar operations in other states around the US.
Customs and Border Protection may want to remember to protect its sensitive facility information. Using basic Google searches, WIRED discovered flashcards made by users of the online learning platform Quizlet that contained gate codes to CBP facilities and more.
In a rare move, Apple this week released “backported” patches for iOS 18 to protect millions of people still using the older operating system from the DarkSword hacking technique that was found in use in the wild. Discovered in March, DarkSword allows attackers to infect iPhones that simply visit a website loaded with the takeover tools embedded in it. Apple initially pushed users to update to the current version of its operating system, iOS 26, but ultimately issued the iOS 18 patches after DarkSword continued to spread.
The US-Israel war with Iran careened into its second month this week, with Iran threatening to launch attacks against more than a dozen US companies, including tech giants like Apple, Google, and Microsoft, which have offices and data centers in the Gulf region. The deadly conflict, which has no clear end in sight, continues to wreak havoc on the global economy as shipping crews remain stranded in the Strait of Hormuz, a key trade route. Meanwhile, some are beginning to wonder what could happen if US strikes cause real damage to Iran’s nuclear facilities.
And that’s not all! Each week, we round up the security and privacy news we didn’t cover in depth ourselves. Click the headlines to read the full stories. And stay safe out there.
Hackers Are Posting the Claude Code Leak With Bonus Malware
Earlier this week, a security researcher flagged that Anthropic accidentally made the source code for its popular vibe-coding tool, Claude Code, public. Immediately, people began reposting the code on the developer platform GitHub. But beware if you want to try to download some of those repos yourself: BleepingComputer reports that some of the posters are actually hackers who have tucked a piece of infostealer malware into the lines of code.
Anthropic, for its part, has been trying to remove copies of the leak (malware-ridden or not) by issuing copyright takedown notices. The Wall Street Journal reported that the company initially tried to remove more than 8,000 repositories on GitHub but later narrowed that down to 96 copies and adaptations.
This isn't the first time that hackers have capitalized on interest in Claude Code, which requires users who might not be as familiar with their computer's terminal to copy and paste install commands from a website. In March, 404 Media reported that sponsored ads on Google led to sites that were masquerading as official Claude Code installation guides, which directed users to run a command that would actually download malware.
Hack of FBI Wiretap Tools Are Officially a National Security Risk
The FBI formally classified a recent cyber intrusion into one of its surveillance collection systems as a “major incident” under FISMA—a legal designation reserved for breaches believed to pose serious risks to national security. The determination, reported to Congress earlier this week, is understood to be the first time since at least 2020 that the bureau has declared a major incident on its own systems. Politico, citing two unnamed senior Trump administration officials, reported that China is believed to be behind the intrusion. If confirmed, the breach could mark a significant counterintelligence failure for the FBI.
The FBI said it detected “suspicious activities” on its networks in February. In a notice to Congress on March 4, reviewed by Politico, the bureau said the compromised systems were unclassified and held “returns from legal process,” citing, as examples, phone and internet metadata collected under court orders and personal information “pertaining to subjects of FBI investigations.” The intruders reportedly gained access through a commercial internet service provider, an approach the FBI characterized as reflecting “sophisticated tactics.” In its only public statement, the bureau said it had deployed “all technical capabilities to respond.”
The breach adds to what has become a pattern of hackers, most if not all foreign, penetrating the FBI's own systems and surveillance infrastructure. In 2023, a foreign hacker accessed files from the bureau's Epstein investigation through an exposed forensic lab server. Last month, Iranian-linked hackers compromised FBI Director Kash Patel's personal email. The Salt Typhoon campaign, uncovered in 2024, saw Chinese hackers burrow into at least eight domestic telecom and internet service providers—exploiting the carrier side of the same surveillance infrastructure believed to be at issue in the current breach. The FBI acknowledged last year that Salt Typhoon had compromised at least 200 companies across 80 countries, and researchers said it showed no signs of slowing down.
How a 22-Year-Old College Student Helped Take Down a Record-Breaking Botnet
Two weeks ago, US law enforcement announced a landmark takedown of four interrelated botnets—massive collections of computers hijacked with malware to do a hacker’s bidding—that were known by the names Aisuru, Kimwolf, JackSkid, and Mossad. The Aisuru and Kimwolf botnets in particular had carried out some of the biggest so-called distributed denial-of-service cyberattacks in history, using hordes of hacked internet-of-things devices to bombard victims with junk traffic.
Now The Wall Street Journal has published a detailed look at an unlikely player in the investigation of those botnets, 22-year-old Benjamin Brundage, a student at the Rochester Institute of Technology. Brundage obsessively tracked the Kimwolf botnet, which he would learn had infected home networks around the world via devices that act as “residential proxies,” essentially offering backdoors into those networks. Brundage went so far as to lurk on Discord and chat with people he suspected had insider information on the hacking campaign, learning key technical clues that he shared with law enforcement. Along with Brundage’s story, the Journal also offered a helpful guide to help determine whether your home network is vulnerable via residential proxy devices and how to protect yourself.
$280 Million Stolen From Drift Crypto Platform, Likely by North Korean Hackers
Given the rate at which the cryptocurrency industry’s insecurity has funded the authoritarian regime of Kim Jong Un in recent years, 2026 was overdue for a large-scale North Korean crypto theft. Now, the decentralized finance platform Drift has conceded that $280 million was stolen from the company in a cybersecurity breach. Crypto-tracing firm Elliptic pointed the finger at North Korean hackers for the intrusion based on clues in their interactions with the blockchains of the stolen crypto as well as their “laundering methodologies and network-level indicators.” In total, Elliptic says that North Korean hackers have stolen close to $300 million this year, the vast majority of which was taken in this latest theft. As huge as that heist may be, the country’s hackers still aren’t quite on track to beat the $2 billion in crypto they stole in total last year.
Cisco Source Code Stolen in Software Supply Chain Breach Spree
Cybersecurity news outlet Bleeping Computer reported this week that Cisco had been the latest victim of a software supply chain hacking spree, which has now resulted in the theft of portions of the company’s source code and that of some of its customers. The breach appears to be the work of the TeamPCP hacker group, which has compromised multiple pieces of security software with its own malicious code, then used their access from that malware to steal user credentials. In this case, Cisco’s credentials were reportedly stolen via the compromise of the vulnerability scanner software Trivy, which then allowed the hackers to access Cisco’s developer environments. The Cisco breach is just the most recent in a string of supply chain attacks that TeamPCP has carried out to spread its infostealer malware, including via the LiteLLM AI software and the security software CheckMarx.