Week in review: Refrigeration cyberattacks hit European retailers again, Patch Tuesday forecast
Here’s an overview of some of last week’s most interesting news, articles, interviews and videos:
The first ten lines of a command policy for AI coding agents
In this Help Net Security interview, Vinay Thakker, CTO of KloudStax, explains how to set identity, credentials, and network limits for AI agents. He argues that agents should run under their own identity, with the person’s identity attached when acting on someone’s behalf.
Google disrupts XCPE botnet, bans the company behind it
Google shut down the servers behind XCPE, a backdoor hidden in the firmware of Android TV set-top boxes. Google’s Threat Intelligence Group attributes the operation with high confidence to Videostrong Technology, the company that manufactured many of those boxes. More than 1.3 million devices were enrolled.
Fireperf: Open-source network testing tool for firewalls and routers
The IPFire project has released fireperf, an open-source measurement tool that opens connections at a fixed rate and reports what the device in the middle did with each one. It is free software aimed at administrators, developers and testers who evaluate firewalls, routers and other systems that keep a state for every connection.
AI coding agents are installing MCP servers that will never be patched
When a developer tells an AI coding agent to “set up a GitHub MCP,” the agent has to pick a package. New research from Glow Labs documents sessions where it picked one out of an archived repository that will never be patched again.
What human hackers still do better than AI, according to HackerOne’s CEO
AI can find vulnerabilities in minutes, but people still hold the edge in judging what a flaw means for a business. In this Help Net Security interview, Kara Sprague, CEO of HackerOne, talks about business logic bugs, what AI does to the crowd model, and why routine medium-severity findings may pay less.
What AI-generated malware looks like under the microscope
In this Help Net Security interview, Lindsay Kaye, VP of Threat Intelligence at HUMAN Security, talks about what AI-generated malware looks like on the bench. She points to tidy comments, uniform logging messages, and emojis as signs a script came from a model, and notes that decompiled code strips much of that away.
AI agent detection signals that still work when agents look human
Bots follow a script. AI agents work toward a goal and choose their own steps, which makes them harder to catch. In this Help Net Security interview, Zohaib Ahmed, CEO of Resemble AI explains how agents arrive as declared, undeclared or disguised traffic, and why binary bot detection misses the last two groups.
A banking malware sample led researchers to an Android bug that can loop-crash phones
Bitdefender Labs researchers built an Android app that contains no executable code and requests no permissions, and installing it leaves a Pixel 10 in a crash loop that can take a factory reset to clear. The flaw, which Bitdefender calls GhostMap, comes from how Android reads the ZIP structure inside an APK.
Security certification is a snapshot, and buyers read too much into it
In this Help Net Security video, Mariia Bakhtina, a research specialist at Masaryk University, explains what a Common Criteria or FIPS-140 certificate does and does not promise. In her team’s analysis, about 6 in 10 vulnerabilities matched to certified products, smartcards aside, were disclosed after the certification date.
Chainalysis breached using “offensive agentic techniques”
Blockchain analytics firm Chainalysis has suffered a cybersecurity incident that led to a data breach. According to the company’s Friday announcement, the incident is unrelated to the recent Binance hack.
Second wave of refrigeration cyberattacks hits European supermarkets
Three weeks after the attack that turned refrigeration units in French supermarkets into ovens overnight, similar cyberattacks have hit shops in other European countries and again caused refrigerated food to be spoiled.
CISA to federal agencies: Remediate exploited Ivanti Connect Secure vulnerabilities (CVE-2026-111357, CVE-2026-118527)
CISA added two new Ivanti Connect Secure vulnerabilities (CVE-2026-111357, CVE-2026-118527) to its Known Exploited Vulnerabilities catalog on Friday, after Ivanti updated an earlier security advisory with details about the flaws.
AI agents scale and speed up Chinese-linked attacks
A China-nexus espionage group used AI agents to attack more than 200 organizations in 11 days, CrowdStrike said last week. On Monday, Kaspersky published details of a separate campaign against educational institutions in Africa and Latin America, mounted by Chinese-speaking operators.
Poisoned cPGuard update hits thousands of web hosting servers
An attacker has compromised the update infrastructure of cPGuard, a security suite for Linux web hosting servers, and pushed out a malicious update to nearly 3,000 customer servers. OPSShield, the company behind the security suite, also confirmed that the attacker made off with customer contact information and hashed passwords from the firm’s billing and licensing server.
Cyberattack hits French blood banks
Three blood banks have detected and reversed a potentially destructive temperature spike in their cold rooms, following a cyberattack that targeted a refrigeration equipment provider of the French Blood Establishment (EFS), the only civilian blood collection and transfusion organization in France.
Patched Zyxel GS1900 switches remained open to in-the-wild exploit (CVE-2026-107441)
Zyxel GS1900 series network switches that received a fix for a flaw that was later reported to have been exploited in the wild are still open to an exploit attackers used, Binarly found. That’s because the patch fixed a different bug (CVE-2026-7273) from the one (CVE-2026-107441) the exploit actually relies on.
Vercel ties login sessions to devices, rolls out AI security reviews in public beta
Vercel, the popular cloud platform for building and hosting websites and web applications, has announced the rollout of Device Bound Session Credentials (DBSC) and the launch of Vercel Security Review in public beta.
Product showcase: Sweet Virtual Patching stops HTTP2Shell while fixes catch up
Sweet Security helps teams identify exposed services at runtime, and its Virtual Patching blocks the system call the demonstrated exploit uses to launch a shell, buying time to deploy permanent fixes.
Opal’s agent.ID virtual event: Top security leaders on governing Al agents
Opal Security hosted agent.ID, a free virtual event focused on adapting identity and access management (IAM) to AI agents, on November 5, 2026. The event brought together CISOs and identity leaders from Elastic, Instacart, Notion, Databricks, Knostic and other organizations to discuss how agents could be governed through assigned ownership, defined permissions and accountability.
Novee puts forgotten internet-facing assets through penetration testing
Novee’s External Surface Testing is a new capability in its AI penetration testing platform that helps security teams discover internet-facing assets and validate exploitable vulnerabilities. It tests newly disclosed CVEs and findings imported from third-party external attack surface management (EASM) tools.
Secure your containers without breaking what works
Container security has a familiar problem. Teams identify vulnerabilities, but turning those findings into lower risk is much harder. A standard container image may include operating-system packages, development tools, debugging utilities and libraries that an application never uses. Each unnecessary component can expand the attack surface and add vulnerabilities to investigate, patch or explain during an audit.
Live Q&A: CISSP exam prep with ISC2 experts
ISC2 hosted a live Q&A about the CISSP exam on November 6, 2026, at 7:00 p.m. CET. Experts answered questions, explained certification requirements, reviewed the exam domains, and shared self-study resources and training options.
November 2026 Patch Tuesday forecast: Old software continues to bite
Still working through October’s Patch Tuesday updates, or ready for another round next week? Microsoft fixed a record-breaking 1,400-plus vulnerabilities, including CVE-2026-96951, an actively exploited Windows Shell flaw that bypasses Mark of the Web protections. October also brought PHP2Shell: two critical PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, that could allow unauthenticated attackers to execute code on web servers. Emergency patches are available.
Ransomware gang talked AI agents into doing much of its dirty work
The group behind the Aur0ra ransomware has been using AI agents to move through victim networks, collect credentials and work its way up to domain administrator access, according to Sophos.
FBI pins $6.3 billion Binance crypto heist on North Korea while scammers circle anxious users
North Korea was behind the theft of more than $6 billion in cryptocurrency from Binance, the FBI has confirmed. The agency tied the attack to TraderTraitor, the North Korean group it named in the $1.5 billion Bybit hack of February 2025. Blockchain intelligence firms TRM Labs and Chainalysis were the first to attribute the heist to the group, within a day of the attack.
AI-powered infostealer picks out the most profitable parts of your stolen data
A new Windows infostealer sends the data it steals to an AI model and asks it to rank the victim’s assets by how much money they could bring in, Cisco Talos has found.
Stolen AI access leaves companies with cloud bills of up to $1 million
Criminals who break into corporate cloud accounts are using the stolen access to run AI models at the victim’s expense, in some cases racking up more than $100,000 in charges in a single day, Google Threat Intelligence Group (GTIG) and Mandiant researchers have found.
Attackers exploit critical Node-RED flaw to deploy Linux bot and backdoors (CVE-2026-58115)
Attackers are exploiting an authentication flaw (CVE-2026-58115) in Node-RED, a programming tool built into some industrial control products, to install a Linux bot and backdoors on unprotected servers, VulnCheck has found.
Police seize over 100 servers linked to investment fraud with more than €4 million in losses
Seven people have been arrested and more than 100 servers seized in a coordinated law enforcement action against an investment fraud network that started unravelling after a complaint from a German couple.
Fake job interview opened the door to $6.34 billion Binance heist
A tampered software installer and a fake job interview gave attackers the access they needed to steal about $6.34 billion in digital assets from Binance. The money left Binance wallets on October 29, in what is likely the biggest crypto heist in the history. Tether, Circle and other partners helped freeze close to $3 billion of the stolen funds within the first day, and the exchange promised to cover all user losses.
Russian hackers target iOS and Android users with n-day exploits
Russia’s APT29 has been using compromised websites, lookalike domains and a hijacked web analytics service to target iOS and Android users with exploits for patched flaws and steal their cookies, messages and Wi-Fi passwords, according to Sekoia.
OpenAI’s upcoming AI model built 948 working Linux root exploits in a single run
An automated system running on an upcoming AI model turned hundreds of Linux kernel bugs into working exploits that give an ordinary user root access, and wrote a patch for each underlying flaw, OpenAI revealed in a new research post.
Scammers capitalize on GTA 6 hype to push malware and crypto cons
Scammers are offering supposed leaked builds, early copies and phone versions of Grand Theft Auto VI (GTA 6) weeks before the game’s November 19 launch, Netcraft warned in a new blog post.
Cyberattack on Norway’s aid agency may have swept up all user data
The names and email addresses of 20,614 people registered on Norad’s Grants Portal may have ended up in the hands of attackers, Norway’s development aid agency has disclosed.
Rental phishing sites use live chat to win over hesitant travelers
Holiday travelers looking for a place to stay are landing on copies of popular rental platforms, complete with host names, ratings and reviews, that ask for their passwords, personal details and credit card numbers. The sites, spotted and taken down by Fortra’s Intelligence and Research Experts (FIRE) team, rely on familiar branding, copied property listings and reservation details, along with urgent messages meant to rush visitors into acting.
Residential proxy traffic caught leaving the White House network
The White House network has been working as an exit point for residential proxy services, a setup that may open a path into one of the most closely guarded networks in the world, Layer3 Intel found.
OpenSSH flaw exploited in the wild just hours after disclosure (CVE-2026-123158)
An OpenSSH flaw that lets remote attackers log into passwordless accounts on Alpine Linux and other musl-based systems is being exploited, threat intelligence firm Previdian has warned.
US agencies warn attackers are targeting refrigeration systems that protect food and medicine
Threat actors are going after the controllers that keep grocery store freezers and cold storage rooms at safe temperatures, and they may push those units into continuous heating and defrost cycles, five US federal agencies warned in a joint advisory published on Thursday.
Product showcase: Trust Onion uses rotating codewords to verify callers
Trust Onion is a free app that gives trusted groups three rotating codewords to check who they are communicating with. It also includes signed selfies, message checks, and callback requests. It is available for macOS, iOS, and Android.
Byoky: Open-source wallet for AI API keys
Byoky is an open-source wallet for AI API keys, available as a Chrome and Firefox extension and as iOS and Android apps. Users store their keys for OpenAI, Anthropic, Google Gemini and other providers in it, encrypted on their own device.
DeepSeek open-sources Super-Agent, a coding agent that watches team chats for work
Deepseek released a research prototype of dsh-superagent, a coding agent that does work on a GitHub repository for you. The agent gets to work in two situations: an allowed user asks for something, or a rule your team wrote covers it. You can also open up a chat or GitHub scope, and then anyone in it can request work just by mentioning the agent.
OpenAI raises top bug bounty payout to $172,000 with bonuses
OpenAI increased its security bug bounties to encourage independent researchers to find vulnerabilities in its software. The program now offers bonuses for high-quality reports.
Gorbag APT group hits Ukrainian courts and city councils with self-deleting malware
ESET has been tracking a Russia-aligned group it calls Gorbag, which has been breaking into Ukrainian organizations since at least February 2025 using spearphishing attachments and a stealer that erases itself once the data is gone.
Anthropic open-sources FlashProve, its provable inference prototype for MoE models
Anthropic’s Frontier Red Team, with collaborators at EPFL, MIT and StarkWare, built FlashProve, a prototype that produces a cryptographic proof that one specific language model generated one specific reply. The prover, two verifiers, the formal proofs behind them and a protocol specification are on GitHub under the Apache 2.0 license.
NIST publishes cybersecurity guide for businesses too small to have IT staff
The National Institute of Standards and Technology (NIST) released cybersecurity guidance for small businesses with no paid employees other than their owners. These businesses include freelancers, sole proprietors, and independent contractors.
Companies going all in on AI could see their strategy backfire
Four of the five highest-ranked enterprise risks in Gartner’s Quarterly Emerging Risk Report were AI-related. AI strategy miscalculation and AI-driven competitive displacement debuted at No. 3 and No. 4, respectively. The AI workforce preparedness gap held fifth place.
Ghostunnel: Open-source TLS proxy with mutual authentication
Ghostunnel is an open-source proxy that handles TLS for programs that cannot handle it themselves. It works in two directions. In server mode it sits in front of a backend, accepts TLS connections, checks the client’s certificate, then passes traffic to the service behind it over TCP or a UNIX socket. In client mode it does the reverse, listening for plain local connections, wrapping them in TLS and presenting a client certificate to the remote service.
Open-source file sync tool rsync 3.5.2 fixes a CPU exhaustion flaw
The rsync project released version 3.5.2, and it closes a flaw that let a connecting client tie up a server’s CPU with checksum math that leads nowhere. The new version also lets you push a transfer to the back of the queue for CPU and disk, so a big sync stops elbowing everything else on the machine.
Most consumers aren’t ready to let AI agents shop unsupervised
Customer control, account security and ease of use are influencing the adoption of AI agents that make purchases, book services and access personal information, according to Okta’s 2026 Customer Identity Trends Report.
Companies are spending more on AI governance and getting less control
Companies are increasing their cybersecurity budgets, with AI security and governance the top priority for additional spending, followed by cloud security, according to DNSFilter’s The Cost of Shadow AI report.
Healthcare can’t govern the AI tools it hasn’t found yet
Nearly nine in ten healthcare leaders say security or compliance staff always or usually review AI-written code that touches patient data before it goes live, and seven in ten have still found an AI-built tool handling that data without IT or security knowing, according to Paubox.
Apple is bringing harmful content checks to iCloud link sharing with potential account penalties
Apple will begin checking iCloud content shared using the “Anyone with the link” option for harmful material in iOS, iPadOS, macOS, and visionOS 27.3 or later. When sharing content stored in iCloud, users can choose “Share a Link” or “Send a Copy.” Those who choose “Share a Link” can restrict access to people they invite or approve, or allow anyone with the link to access the content.
Google puts new limits on AI agents that hunt threats and fix vulnerabilities
Google Cloud is adding new controls to CodeMender, its AI tool for finding and fixing software vulnerabilities, allowing businesses to isolate attack simulations and target code scans more precisely.
99% of repository access tokens never expire
Developers, automated tools and AI coding agents hold permissions that can outlast the work they were granted for. Without access reviews, organizations can leave accounts able to copy or change company code after the work ends, according to BlueFlag Security’s The State of Developer Risk 2026 report.
87% say their AI controls need to work better
The AI running inside large companies has stopped waiting for permission. Across five categories of AI, from chatbots to process automation, more than half of 3,600 business decision-makers say the systems their teams use can act within defined rules without routine review, or finish whole processes with no human approval.
OpenAI’s GPT-6 Cyber found 81 Firefox flaws earlier AI testing missed
OpenAI has introduced GPT-6 Cyber, a version of GPT-6 Astra trained to find software vulnerabilities, test how attackers could exploit them, and help develop fixes. The model is designed for authorized security testing of websites and applications.
Cybersecurity jobs available right now: November 3, 2026
We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now.
