Last week's question was why the software ecosystem keeps shipping holes and handing the cleanup bill to operational teams. This week npm answered, at least partially. npm v12 will block automatic code execution during install by default — no more preinstall scripts running silently, no more Git dependencies or URL-based packages pulling in whatever they feel like. Developers will have to explicitly opt in. It's the right call, it's what the supply chain attack surface has been screaming for across months of CanisterWorm, Shai-Hulud, IronWorm, and Megalodon campaigns, and it arrives roughly four years after the attack pattern became impossible to ignore.
The malware section this week is, as ever, the context that makes the fix legible. Nineteen PyPI packages trojaned via .pth startup hooks. A WinRAR flaw from last year still fueling active campaigns against Ukrainian organizations. TeamPCP back with CanisterWorm. The backlog of techniques that predate npm v12 isn't going anywhere — and the install-time execution block doesn't touch the packages already in production, the developers who won't upgrade immediately, or the registries that aren't npm. It's a meaningful fix to a well-understood problem. It's also, by the industry's own timeline, a very belated one.
Table of Contents
🔓 BREACHES & SECURITY INCIDENTS
🇮🇷 🇺🇸 Iran-linked hacker group Handala says it breached California Water Service and posted 5 GB of stolen data. The leak includes customer billing records, account details, and admin credentials for an RTKBase GNSS system. Security firm Dataminr warns the group could escalate to destructive attacks and urges credential rotation and audits.
🇩🇰 Pharma giant Novo Nordisk reported a data breach exposing pseudonymized clinical trial patient data and personal details of some healthcare professionals. The company says names were not linked to the patient data but warned HCPs to watch for phishing and fraudulent contact. Novo Nordisk has taken affected systems offline and is working with cybersecurity experts to investigate.
🔓 Cybercrime group ShinyHunters says it breached Oracle PeopleSoft servers at over 100 organizations, many universities. Stolen data reportedly includes student addresses, phone numbers, emails, birthdates, and other sensitive records. The group targets common software vulnerabilities to compromise many victims at once.
🩹 ServiceNow fixed a security flaw that let unauthenticated attackers query customer data via a vulnerable API. Impacted customers were told via support cases and a hidden bulletin; ServiceNow is still investigating what was accessed. Admins reported the issue tied to a REST endpoint and shared indicators of compromise.
🇫🇷 Hackers used a hijacked user account to breach Tchap, France’s government encrypted messaging platform — DINUM says the account was blocked and an investigation is ongoing, and users were warned that public rooms are not encrypted. A threat actor claims to have stolen ~13.5 GB of data, thousands of messages, and tens of thousands of account details.
🇭🇰 SoFi Hong Kong says hackers accessed a database at a third-party vendor and a data breach occurred. The company discovered the incident on April 30, 2026, and is investigating with a cybersecurity firm. Customers are urged to watch for phishing, secure their accounts, and contact SoFi for help.
🇬🇧 Oxford University disclosed a data breach after its third-party CareerConnect platform was hacked on May 28. Attackers accessed names, emails, and encrypted passwords for users who do not use Single Sign-On. The university says no university systems or financial data appear compromised but warns of possible phishing.
→ More breaches:
🥷🏻 CYBERCRIME, CYBER ESPIONAGE, APT’s
🔞 ❌ U.S., French, and Italian authorities seized domains for CFAKE.com and SOCFAKE.com for hosting thousands of nonconsensual deepfake porn images and videos. The sites let users search disturbing categories like “rape” and “degradation” and targeted public figures and private women. A French suspect was arrested, assets seized, and prosecutors say the sites involved hundreds of thousands of images and thousands of videos.
🎣 ❌ INTERPOL led Operation Ramz and shut down Sniper Dz, a decade-old phishing-as-a-service platform — Authorities from 13 MENA countries made 201 arrests, including the platform’s main developer, and seized its servers and tools. Sniper Dz had hosted thousands of phishing sites targeting major brands and harvested credentials and other data for fraud.
🇺🇸 ⚖ 🇷🇺 U.S. prosecutors charged Russian citizen Denis Obrezko for helping a large cyber-espionage campaign tied to the Russia-linked group Void Blizzard. Investigators say he bought servers and domains used to break into companies, schools, and other organizations. The group used stolen session tokens and simple proxy tricks to steal emails, files, and access to cloud accounts.
🇺🇸 ❌ 🇨🇳 The FBI seized 13 fake websites that posed as consulting firms to lure U.S. workers with security clearances. Officials say the sites were used by operatives tied to Chinese intelligence to recruit people and buy sensitive information. China’s embassy denies the allegations.
🇻🇳 🪷 Vietnam-linked OceanLotus used the SPECTRALVIPER backdoor in two campaigns that targeted a Vietnamese construction firm and stock investors via a compromised FireAnt Metakit updater. The attackers delivered SPECTRALVIPER through DLL side-loading and a malicious update, enabling data theft, lateral movement, and remote control. ESET warns this shows a shift toward more selective domestic espionage by the long-running group.
😶 Silent Ransom Group (SRG) uses social engineering and vishing to get into victims’ networks, often targeting law firms. They steal data, pressure victims with extortion emails, and sometimes use in-person USB drops. Resecurity found SRG hides its infrastructure with a global fast flux network of infected routers and IoT devices.
🗓️ {Cyber,Info}Sec Events — A community-maintained list of infosec conferences worldwide. Subscribe to the ICS calendar feed to get events straight into your calendar, or follow @[email protected] on Mastodon for weekly digests. Contributions and ⭐ welcome!
👨🏻⚖️ 👀 GOVERNMENT, POLITICS, AND PRIVACY
🇺🇸 ⚖ Meta says it found a spearphishing campaign linked to NSO Group that violated a court injunction — Meta filed a contempt-of-court complaint after disrupting the attacks and removing fake WhatsApp accounts. Critics say NSO’s behavior supports keeping it on the U.S. sanctions Entity List.
🇺🇸 📍 Massachusetts lawmakers passed a Consumer Data Privacy Act giving residents new rights to access and delete personal data. The bill bans companies from selling precise location data and other sensitive information without consent. It now goes to the Senate and the governor and is expected to become law.
🦠 MALWARE & THREATS
🆕 📦 GitHub says npm v12 will block automatic code execution and remote dependency sources during npm install unless explicitly approved. This stops install scripts, Git dependencies, and URL-based packages from running by default to reduce supply-chain attacks. Developers must opt in for these behaviors and can test with npm 11.16.0 before upgrading.
❌ Microsoft temporarily removed 73 GitHub repositories after detecting possible malicious content linked to a Miasma/Shai-Hulud supply‑chain campaign. The takedown briefly broke Azure Functions deployment workflows, but the repos were restored and deemed clean. Microsoft is investigating and notifying affected customers.
🇷🇺 🇺🇦 Two Russia-aligned groups used a patched WinRAR flaw (CVE-2025-8088) to infect Ukrainian organizations with data-stealing malware. They delivered payloads via crafted RAR archives and ADS files, then used loaders to steal browser data, passwords, and documents. Attackers shifted to dedicated C2 servers and long-running infection chains to maintain stealthy access.
🐍 Hackers trojanized 19 popular PyPI packages in a Shai-Hulud supply-chain attack to steal developer secrets. The malicious packages used .pth startup hooks to run an obfuscated JavaScript payload via the Bun runtime and exfiltrated tokens, cloud credentials, SSH keys, and more. Socket urges affected users to rotate secrets, restore from clean backups, and watch for .pth hooks or unexpected Bun downloads.
🏦 New NFCShare Android malware is being spread as fake bank app updates on GitHub. It tricks users into tapping cards to their phones, steals card data and PINs via NFC, and sends them to attackers. Users should only install banking apps from Google Play and ignore unexpected NFC verification prompts.
🇨🇳 🔙 🚪 A China-linked group called VerdantBamboo deployed a BSD variant of the BRICKSTORM backdoor plus PLENET and AGENTPSD to compromise Linux appliances and NAS devices. They gained access by breaching an MSP and abusing VPN/firewall credentials to move laterally and blend in with normal traffic. The actor uses tailored implants, living-off-the-land techniques, and careful operational security to maintain long-term access.
🤖 🧰 AI, CRYPTO, TECH & TOOLS
❌ Law enforcement dismantled AudiA6, a crypto-laundering service that moved over $380 million for ransomware and other cybercrimes. Investigators from 11 countries arrested two alleged admins, seized domains, vehicles, crypto, and KYC records tied to thousands of mule accounts. The platform and its forum Dark2Web now show seizure notices and the suspects face heavy prison time.
🆕 Anthropic released Claude Fable 5, a powerful Mythos-class AI with built-in safety blocks that prevent use in high-risk areas like cybersecurity. In sensitive queries the model automatically falls back to the less capable Claude Opus 4.8, and tests and external red-teaming found no universal jailbreaks. Trusted partners in Project Glasswing get upgraded Mythos access, and both Fable 5 and Mythos 5 are available via API with published pricing.
Anthropic says claims that Claude Fable 5 was jailbroken are false. The company says its separate classifier safeguards still block dangerous outputs even if the model is coaxed to respond. Reviewers found no evidence the system was bypassed to produce harmful, nonpublic guidance.
🇺🇸 The U.S. government ordered Anthropic to shut down its two most powerful AI models, Claude Fable 5 and Claude Mythos 5, over national security concerns. Anthropic says the move is wrong and that any jailbreak risk is narrow and already present in other models. Critics note the company’s warnings about Mythos may have drawn the government’s scrutiny.
🔐 OpenAI is expanding two ChatGPT security controls to help protect accounts and data. Lockdown Mode limits features and outbound requests to reduce data exfiltration from prompt injections. Active Sessions shows where you are signed in, and Advanced Account Security adds passkeys and tighter sign-in protections.
🐛 🧠 VULNERABILITIES, RESEARCH, AND THREAT INTELLIGENCE
➝ From the Patching Department:
💥 A high-severity unpatched flaw (CVE-2026-5027) in Langflow lets attackers write files anywhere via path traversal and can lead to unauthenticated remote code execution. The bug is exploitable without credentials because Langflow enables auto-login by default, and attacks are already observed in the wild. Thousands of Langflow instances are exposed online, raising risk as attackers increasingly target AI development tools.
🐧 A one-character bug in Linux nf_tables (CVE-2026-23111) lets an unprivileged local user escalate to root and break out of containers. Public working exploits were published in April and June, and the upstream one-line fix shipped February 5, 2026. Update affected kernels and reboot, and restrict unprivileged user namespaces until patches are applied.
💥 A high-severity command injection bug in BerriAI LiteLLM (CVE-2026-42271) is being actively exploited, letting authenticated users run commands on the host. Researchers say it can be chained with a Starlette host header bypass (CVE-2026-48710) to achieve unauthenticated remote code execution. Users should update LiteLLM to 1.83.7+ and Starlette to 1.0.1+ or apply immediate mitigations like blocking the test endpoints.
🩹 Check Point patched a critical VPN flaw (CVE-2026-50751) that lets remote attackers bypass authentication on devices using the deprecated IKEv1 protocol. The bug was exploited in zero-day attacks affecting a few dozen organizations, with at least one incident tied to the Qilin ransomware group. Check Point also found a second IKEv1-related certificate validation bug (CVE-2026-50752) and urged immediate updates and mitigations.
🛰️ ICS, OT & IoT
🩹 ICS Patch Tuesday — Siemens, Schneider Electric, and Phoenix Contact released security advisories fixing multiple industrial control system vulnerabilities this month. Siemens patched flaws including remote code execution and information leaks, Schneider fixed DoS, credential, and information disclosure issues, and Phoenix Contact addressed an unauthenticated log download bug. CISA and VDE CERT also issued related notices while other vendors like Rockwell, ABB, and Mitsubishi posted updates or enhancements.
🦠 C0XMO is a new, modular Gafgyt-based botnet that exploits a DD-WRT router flaw to infect many device types and CPU architectures. It spreads by scanning, brute-forcing credentials, installing persistent binaries, and killing rival malware. Once connected to its C2, it can run 19 DDoS attack methods and be updated remotely.
💬 CONNECT
Follow me on Mastodon for quick daily updates and bite-sized content.
Prefer using an RSS feed? Add Infosec MASHUP to your feed here.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
Thanks for reading today’s newsletter, and if you're enjoying it and want to support my work, you can buy me a coffee ☕ over at https://www.buymeacoffee.com/0x58
See you next time!
-X.
