U.S. officials suspect Iran-linked hackers are behind a string of cyberattacks that have made freezers malfunction at around six military commissaries, the on-base grocery service members use. According to two people familiar with the matter, the intrusions began in late August. Cyber Command and several Defense Department law enforcement units are investigating, and early indications suggest Iran may be to blame.

How the attackers got in isn’t yet clear, but internet-exposed refrigeration controls are a known weak spot, and the vulnerabilities in them can leave equipment open to sabotage. French authorities are now working with U.S. investigators, after hackers overrode temperature controls on commercial refrigeration at about 350 French establishments in October and forced supermarkets to throw out spoiled food.

It’s a reminder of the unglamorous systems nobody puts on a threat model. If something has an IP address, it has the potential to be compromised, no matter how quotidian it is.
https://www.nextgov.com/cybersecurity/2026/10/officials-suspect-iran-series-cyberattacks-military-base-freezers/416934/

Did someone forward you this? Subscribe to this newsletter.

In other cybersecurity news:

A Mystery 7.72 TB Numbat Checkpoint Claims to Leak Anthropic's Next AI. A Hugging Face user named burner22/numbat has published 772 safetensors shards totaling about 7.72 TB. Its model card describes the files as a mid-pretraining checkpoint from Anthropic’s next “Mythos-class” run, with roughly 14 trillion parameters and the internal codename Numbat. An authentic checkpoint of this scale could expose valuable model weights, architecture choices, and training practices. Anthropic has not publicly commented on the repository, and the available evidence does not establish an Anthropic connection or a data breach. The checkpoint’s provenance remains unverified.
https://alphasignal.ai/news/a-mystery-7-72-tb-numbat-checkpoint-claims-to-leak-anthropic-s-next-ai
https://huggingface.co/burner22/numbat

Binance said Sunday that it has resumed withdrawals of bitcoin and other cryptocurrencies following the record theft of around $6.3 billion in token assets from its hot wallets. “First batch network now opened in phase one of withdrawals is $BTC bitcoin network,” Binance said in an X post. It later said withdrawals have also been restored for the Ethereum Network and BNB Smart Chain.
https://www.cnbc.com/2026/10/31/binance-hack-asset-coverage-withdrawals.html

Earlier this year, sources tell WIRED, a contractor caused a security incident at the Department of Defense by connecting a secure centralized system used by multiple intelligence agencies to their personal account for Claude Code, Anthropic’s AI coding agent. It appears that the agent used by the contractor, who is employed by the global consulting firm Accenture, then sent information about the DOD system, including login credentials, to a high-risk, third-party website. The centralized system contains controlled unclassified information (CUI), which is considered sensitive and not to be released to the public without review. CUI can include information about critical infrastructure, nuclear submarine propulsion, and vulnerabilities to chemical and biological terrorism. Additionally, security experts say, access to the system could allow for the exploitation of even more sensitive ones.
https://www.wired.com/story/a-defense-contractor-gave-their-personal-ai-agent-access-to-a-secure-pentagon-system/

An unprecedented leak of Department of Homeland Security records has exposed details of more than 41,000 employees or contractors who work for the department and its constituent agencies. The records include email addresses, phone numbers, and office addresses, but also references to internal systems and sensitive information on undercover and surveillance programs. In all, the leak provides a map of how DHS is organized and how it operates. The records, which span DHS itself as well as seven component agencies, including Immigration and Customs Enforcement and US Customs and Border Protection, were last updated on August 19, 2026. The transparency nonprofit Distributed Denial of Secrets and cybersecurity researcher and journalist maia arson crimew previewed the data for WIRED, which they say they obtained from a group calling itself the Department of Peace.
https://www.wired.com/story/massive-leak-exposes-41000-homeland-security-workers-including-undercover-ice-agents/

Polymarket resumes trading after $2.65 million exploit, says third-party provider was compromised. Arkham data showed that a wallet address identified as belonging to a suspected attacker holds about $2.65 million in POL after revealing multiple recent USDC.e-to-POL swaps through Relay.
https://www.theblock.co/news/defi/2026-10-29-polymarket-exploit-2-5-million-cancel-only-mode-432158

A North Korea-aligned hacking group has moved beyond fake job interviews and now steals developer credentials to plant malware in trusted software projects. The group, tracked as DeceptiveDevelopment, used the stolen access to change code in public and private repositories between April and September 2026. Every poisoned project then became a new way into the systems of the people who use it. The group has been known for years for tricking developers with fake recruiters and coding tests. In earlier campaigns, it mainly wanted to steal cryptocurrency wallets and browser logins.
https://cybersecuritynews.com/north-korean-hackers-2/amp/

Cursor reported that a tampered copy of Claude Code extension (anthropic.claude-code, versions 2.1.313 and 2.1.312) was available and served in the Cursor extension marketplace for a few hours to attempt a supply-chain attack on Cursor users. Cursor reports that it was live from about 2:27 AM to 1:35 PM Pacific Daylight Time (PDT).
https://www.microsoft.com/en-us/security/blog/2026/10/30/inside-malicious-claude-code-branded-cursor-extension-persistent-ssh-backdoor/

Security researchers at Anthropic have developed an exploit for a critical Linux kernel vulnerability that allows unauthenticated attackers to remotely execute commands as root on servers running Multipath TCP (MPTCP) listeners. The flaw, tracked as CVE-2026-98432, also affects ordinary TCP connections, although remote code execution has not been publicly demonstrated on servers without an MPTCP listener.
https://www.bleepingcomputer.com/news/security/critical-linux-kernel-flaw-allows-remote-root-attacks-on-mptcp-servers/

On Monday, University of New Mexico Health Sciences Center information technology teams identified a cybersecurity threat around 11 a.m. and began containment efforts. From 12:30 to 2:30 p.m., the health system temporarily diverted ambulance traffic to other hospitals in Albuquerque and Santa Fe, while outpatient surgeries and appointments were suspended for the remainder of the day as teams worked to protect the network. Notably, UNM Hospital is New Mexico’s only Level I trauma center.
https://gizmodo.com/hacker-group-claims-unm-hospital-cyberattack-then-apologizes-and-offers-patients-2500-2000832169

The Ransomnews research team counted 484,142 publicly listable cloud storage buckets holding 18.05 billion files in October 2026, using data from GrayhatWarfare, the search engine that indexes open cloud storage. Amazon S3 hosts 72% of them, but Azure Blob is 2.5 to 3.6 times over-represented among buckets named for backups, exports and payroll. Open buckets also list 321,058 credential and key files, and 67% of the exposed .env files were last written this year.
https://ransomnews.com/cloud-exposure-index-2026/

On 30 October 2026, Wiz counted nine AI developer platforms that disclosed security incidents between 6 and 30 October: Fireworks AI, Andromeda, Zeabur, Lightning AI, OpenCode, Railway, Warp, CommandCode and Northflank (Wiz). Wiz reported credential access as the entry point at Fireworks, Zeabur and Lightning AI, and Railway’s preliminary finding as “privilege escalation in our build infrastructure”. Not every incident was the same, and not every one exposed stored secrets.
https://vulnpocalypse.dev/advisories/black-october-ai-platform-breaches/

Threat actors can sabotage AI models by tampering with their underlying infrastructure, such as the firmware and software powering the computers that run the models, the security firm Eclypsium said in a report published on Thursday. “A subtle runtime change could bias the model toward weakening certificate validation, selecting an unsafe dependency, or generating a specific insecure pattern,” according to the report. Eclypsium’s findings highlight an underappreciated area of risk that requires safeguards on the part of frontier AI labs and oversight by businesses using them.
https://www.cybersecuritydive.com/news/ai-hardware-model-tampering-eclypsium/873177/

A critical certificate validation flaw in LXD and Incus could let remote attackers bypass login checks and take control of clusters and managed virtual machines. The weakness affects the HTTPS management listener, placing exposed Linux systems at risk without requiring a password, stolen private key, or user interaction. Incus rates its flaw critical, with a CVSS score of 10.0.
https://cybersecuritynews.com/certificate-validation-flaw-in-lxd-and-incus/

A network of 319 browser extensions has been linked to search hijacking affecting roughly 13.4 million users over a decade. Marketed as PDF converters, manual finders, safe search tools, and AI search helpers, these extensions redirect queries through operator-controlled servers before sending users to Yahoo or Bing. The business model turns ordinary searches into referral income while exposing queries to an intermediary. Similar behavior appeared in an earlier Chrome extension search hijacking investigation involving 23 add-ons, showing how basic browser settings can support large advertising operations.
https://cybersecuritynews.com/319-chrome-extensions/amp/

A hacking group claiming responsibility for displaying pornographic footage on a roadside billboard in Egypt told The Cairo Report that it exploited a software vulnerability affecting digital advertising screens in several countries, exposing what it described as widespread security failures in commercially operated infrastructure.
https://www.cairoreport.org/p/egypt-brazil-iraq-billboard-hack

The U.S. Department of Justice (DOJ) just announced indictments resulting from Operation Ghostbusters, which investigated widespread student loan fraud. According to the agency, bad actors used stolen identities to sign up for student loans and used AI to complete the application, making it appear that the fake student was attending online classes.
https://www.tomshardware.com/tech-industry/cyber-security/ghost-students-using-ai-to-apply-for-student-loans-with-stolen-identities-dojs-operation-ghostbusters-busts-usd200-million-ring-faking-online-college-attendance-unsuspecting-victims-saddled-with-student-debt-they-didnt-sign-up-for

The late-October Synology NAS ransomware reports reveal a patching and recovery problem, but the public evidence does not establish one entry point for every affected device. Synology's October 29 statement points to a vulnerability patched in December 2024. Security firm 360 attributes the incidents it investigated to two flaws disclosed in September 2026. These statements must be examined separately, rather than joined into a confirmed attack chain.
https://shop.zimaspace.com/en-ca/blogs/tech-ai-hub/synology-nas-ransomware-c3rb3r-dsm-security-gap

Hackers are using two powerful iPhone exploit kits, Coruna and DarkSword, to break into phones and steal private data. The victim only has to open a web page. There is no app to install and no prompt to accept. The scale is larger than many people expected. Researchers found 608 domains delivering Coruna-family code. These sites choose an exploit chain based on the iOS version on the phone.
https://cybersecuritynews.com/hackers-use-zero-click-coruna/

A new ClickFix campaign is using fake Cloudflare verification pages to trick both Windows and Mac users into running harmful commands on their own computers. The pages look like the “Verify you are human” checks that many people click through every day. Instead of a quick security check, the visitor ends up starting an infection by hand. The campaign was seen on Wednesday, October 28, 2026, and it runs on legitimate websites that have been compromised. A visitor lands on a normal-looking site, and a hidden script loads the fake Cloudflare page on top of it.
https://cybersecuritynews.com/fake-cloudflare-verification/amp/

Hackers have begun testing a critical Oracle E-Business Suite vulnerability that can let them run commands without logging in. Threat intelligence firm Defused reported exploratory exploit attempts against its decoy systems on October 27, 2026. However, these observations do not confirm successful attacks against customer systems, and Oracle had already released a patch before the activity was detected.
https://cybersecuritynews.com/oracle-ebs-rce-exploited/amp/

Hackers Use Malicious Instagram Ads to Infect iPhones With Spyware Without a Click. A documented attack used a World Cup advertisement to deliver the P7 variant of DarkSword, exposing sensitive information without requiring the victim to tap the ad, download a file, or install an application. The findings show how online advertising can become a delivery channel for mobile attacks. Unlike phishing, these infections do not depend on persuading someone to open a suspicious link.
https://cybersecuritynews.com/malicious-instagram-ads/amp/

Fake PlayStation 5 emulators hosted on GitHub are delivering malware that alters Microsoft Defender settings before launching a suspected data-stealing program. The campaign exploits gamers' willingness to experiment with unofficial software, making malicious downloads look like a normal part of the process. It highlights how familiar project names can earn users' trust before they verify who is behind the files.
https://www.techspot.com/news/114376-gamers-hunting-ps5-emulators-actually-downloading-malware-instead.html

Professional poker is the latest community to reckon with the impact of a high-profile cyberattack. Wired reports that hundreds of online poker players have allegedly been targeted in a malware attack that allowed a cheater to see their screens.
https://www.engadget.com/2301500/hacker-may-have-stolen-millions-in-a-malware-attack-against-online-poker-players/

A 16-Year-Old UEFI Bug Lets One Packet Beat Secure Boot. CVE-2026-97631, a 9.5-rated overflow in EDK II's PXE driver dating to 2010, can let someone on the local network run code before Secure Boot on network-booting machines. Firmware updates are still pending.
https://threatroundup.com/news/edk2-pxe-uefi-flaw-cve-2026-97631/