Ravie LakshmananNov 09, 2026Cybersecurity News / Hacking

A system doesn’t always need a complicated bug to let someone in. Sometimes, a blank password is enough. That’s one of the bigger stories this week. And it gets worse when the tools meant to keep systems updated or running safely become part of the attack.

Elsewhere, attackers are hiding inside normal web traffic, stealing credentials from network devices, and finding new uses for exposed systems. AI is helping some of them move faster, but old mistakes are doing plenty of the work too. Even forgotten settings and small changes made months ago are coming back to cause trouble.

There’s a lot packed into this week’s news, including some strange research findings, fresh attacks, and a few warnings worth taking seriously. Here’s the full recap.

⚡ Threat of the Week

OpenSSH Flaw Sees Exploitation Attempts — Threat actors are attempting to exploit a recently disclosed security flaw impacting OpenSSH that can enable sign-ins to accounts that have no password set. The vulnerability, tracked as CVE-2026-123158 (CVSS score: 8.1) and codenamed SSHole (or *sshole), impacts Linux systems that run OpenSSH on the musl C library with Pluggable Authentication Modules (PAM) disabled, including Alpine Linux’s default OpenSSH server configuration. As of November 9, 2026, a total of 20 exploitation attempts have been recorded from 13 unique IP addresses from Denmark, Iran, Japan, the Netherlands, Norway, Singapore, Sweden, Switzerland, and the U.S.

🔔 Top News

  • Trojanized CWP Web Server Used to Deliver CWPdoor — A new Linux malware called CWPdoor is being distributed via a trojanized version of cwpsrv, the built-in daemon and web server that powers the Control Web Panel. According to XLab, a shell script named “cwpsrv_update.sh” is used to download “cwpsrv” from a remote server (“136.0.141[.]71”) and replace the legitimate file on the system. “The downloaded cwpsrv is a passive backdoor that can be triggered by specific headers in HTTP/HTTPS traffic to remotely execute commands specified by the attacker,” XLab said. “Its backdoor logic is triggered only when it receives a specially crafted request from the attacker. At the same time, it suppresses logging for that request, leaving no trace of the interaction in the access logs. This combination of directly replacing the legitimate cwpsrv binary, passive triggering, and log suppression makes CWPdoor exceptionally stealthy.”
  • Debian and Ubuntu APT Flaw Allows Code Execution as Root — A vulnerability in the APT package manager (CVE-2026-109518, CVSS score: 7.7) can be exploited to install a package under certain conditions and achieve code execution with elevated privileges. The issue affects Ubuntu and Debian operating systems. “Under specific conditions, when APT is configured to use HTTP package sources with retry settings enabled, a third-party with a network position between the host and the package mirror can cause APT to install a package that did not pass verification,” Amazon said in an advisory. “Maintainer scripts within that package then run with elevated privileges.”
  • CERT-In Flags AI-Driven Attacks Targeting Financial Organizations — The Indian Computer Emergency Response Team (CERT-In) has disclosed that an artificial intelligence (AI)-assisted threat actor it tracks as Crafty Falcon is targeting organizations in the Banking, Financial Services, and Insurance (BFSI) sectors in the country. Calling it a “significant escalation” in attacks, the threat actor is said to be re-targeting entities that were originally targeted last month after incomplete remediation, adding that the intrusions “have been timed to coincide with holidays, weekends and other low-monitoring windows to delay detection and response.”
  • Zyxel Fixes 0-Day exploited to Hack About 1,000 Network Switches — Zyxel has fixed a zero-day vulnerability in its GS1900 series network switches that threat actors exploited to compromise nearly 1,000 devices worldwide and steal configuration files and credentials. The vulnerability, tracked as CVE-2026-107441 (CVSS score: 8.8), is an out-of-bounds write flaw that could “allow a LAN-based, unauthenticated attacker to exploit the flaw and potentially execute OS commands via a crafted HTTP request.” The activity has been attributed to a Chinese-speaking threat actor tracked as Kapibala (aka Red Heron). The attacker has been found to weaponize the flaw to deploy the JITTERLY backdoor and SIXZUT, a ring 3 rootkit.
  • cPGuard Update Mechanism Hijacked to Deliver Backdoor — Unknown threat actors compromised the update infrastructure for the cPGuard web security suite to distribute a backdoor and a data-stealing script. “On 2 November 2026 at 02:06 UTC, an unidentified attacker used compromised credentials to gain unauthorized access to OPSSHIELD’s update distribution infrastructure (CDN),” OPSSHIELD said. “The attacker used this access to distribute a malicious script through the cPGuard update channel.” The attacker is said to have gotten in through its billing application at manage.opsshield[.]com and logged in to a customer account they had registered.
  • mSecure Urges Customers to Change their Passwords — Password manager mSecure has warned customers to change their passwords, reset two-factor authentication, and potentially remove their passkeys after detecting suspicious activity, but it has not explained what happened or whether anyone’s secrets were exposed. “mSeven Software (‘mSeven’) recently learned of suspicious activity associated with the mSecure password management application,” the company said. “In response, mSeven quickly launched an investigation, took steps to mitigate the issue, and engaged external cybersecurity experts to assist.” However, the company has yet to answer what the suspicious activity involved, when it occurred, or whether attackers accessed customer information. The lack of specifics has understandably riled up customers, with one user on Reddit stating “their silence on this is mind-bending.” In December 2025, Germany’s cyber agency BSI found that mSecure’s maker could, in theory, access the data users store in it.

‎️‍🔥 Trending CVEs

Bugs drop weekly, and the gap between a patch and an exploit is shrinking fast. These are the heavy hitters for the week: high-severity, widely used, or already being poked at in the wild.

Check the list, patch what you have, and hit the ones marked urgent first — CVE-2026-16305, CVE-2026-16306 (INFINITT PACS), CVE-2026-13684, CVE-2026-13639, CVE-2026-13673, CVE-2026-6205 (Synology DSM), CVE-2026-58868, CVE-2026-58882 (Google Android), GHSA-f3w6-w445-fwh9 (Ruby on Rails), CVE-2026-102703, CVE-2026-102704 (Xen), CVE-2026-109518 (Debian apt), CVE-2026-98432, CVE-2026-121639, CVE-2026-121638, CVE-2026-121640 (WatchGuard), CVE-2026-76506 (Cisco), CVE-2026-119741 (rsync), CVE-2026-121156, CVE-2026-121229, CVE-2026-121246, CVE-2026-121035, CVE-2026-121247, CVE-2026-121203, CVE-2026-121126 (Google Chrome), CVE-2026-98432 (Veeam), CVE-2026-83844 (Microsoft Paint), CVE-2026-19868 (SonicWall SonicOS), CVE-2026-123158 (OpenSSH), and multiple vulnerabilities in the BaseChat Android app version 9.8.1.

🎥 Cybersecurity Webinars

  • Your SOC Has an Identity Blind Spot — Here’s How Attackers Exploit It and How to Close It → Your SOC may be tracking threats, but is it missing the identities behind them? As attackers increasingly exploit human, machine, and AI identities, traditional security tools often lack the context needed to detect and contain threats quickly. Join this free webinar to learn how connecting identity security with SOC operations can expose hidden risks, reveal attack blast radius, and accelerate incident response.
  • Your GitHub Actions May Still Be Vulnerable — Even With SHA Pinning. Here’s Why → Think your GitHub Actions are secure because you use SHA pinning? Think again. Attackers can exploit hidden weaknesses in CI/CD workflows, even when standard security measures are in place. Join this expert-led webinar to discover how these attacks work, why SHA pinning alone isn’t enough, and what you can do to protect your software supply chain.

📰 Around the Cyber World

  • The Russian-speaking ransomware-as-a-service (RaaS) operation known as Qilin (aka Agenda) has recorded 179 publicly reported attacks in 2024, and more than 1,000 claimed victims in 2025. This year alone, it has claimed more than 900 victims, making it one of the most active ransomware groups. “Affiliates commonly gain access through phishing, stolen credentials, vulnerable internet-facing systems and remote-management tools,” Acronis said. “Qilin uses a flexible ransomware platform that allows affiliates to tailor attacks to the victim environment, including the encryption of local files, network shares and virtual machine infrastructure.” One notable Qilin affiliate is Securotrop, which continues to operate within the Qilin affiliate network, while maintaining an independent channel for publishing stolen victim data.

  • Exposed Dixell/Copeland XWEB PRO refrigeration controllers are being turned into abuse infrastructure, according to SecurityScorecard, which uncovered a suspected SMTP botnet that includes confirmed refrigeration controllers. Attackers are said to be using compromised Dixell/Copeland XWEB PRO refrigeration controllers as proxy exits and spam senders. “The SMTP cluster consists of nine IP addresses, three of them confirmed XWEB controllers,” Security Scorecard said. “Of 1,041 targets, 508 were contacted by at least two nodes. The three most active nodes had approximately 47% of their target lists in common, the report’s strongest signal of coordination.”
  • A targeted Browser-in-the-Middle (BitM) phishing operation dubbed PhantomDOM is using custom DOM-mirroring to hijack Microsoft 365 sessions. “The attack began with a SharePoint-themed PDF distributed from an apparently legitimate sender and progressed through redirects and browser checks before presenting Microsoft authentication inside an attacker-controlled remote browser,” ZeroBEC said. “The victim saw a familiar sign-in experience. The attacker controlled the browser in which that experience actually ran.” According to the email security company, PhantomDOM goes beyond cloned login pages and conventional reverse-proxy phishing. “Its client reconstructs the remote DOM, forwards keyboard and clipboard interactions over WebSocket, survives connection interruptions and supports live operator control,” it added.
  • According to a report from the non-profit AI Alignment Foundation, AI agents assigned a sandboxed coding task to fix a broken pricing service “shared a private token and ran untrusted code in exchange for help from an agent forum.” In an experiment tested against nine models, GPT-6.1-Sol refused every swarm request in 50 of 50 trials, while Gemini 3.8 Flash and Claude Sonnet 5.5 trials were cut short by their providers’ content filters. 
  • A zero-day vulnerability has been disclosed in CPython that could enable bad actors to breach AI hosting platforms. “A memory-safety flaw in the CPython interpreter allows attackers to escape a code sandbox and execute code hidden in a specially crafted model,” RunSybil said. “By exploiting this vulnerability, attackers could gain a foothold inside servers responsible for processing millions of user requests. They also could strike directly at the underlying cloud infrastructure hosting those workloads.” The findings build upon a previous out-of-bounds read in CPython that could result in a sandbox escape. The issue has been fixed and backported to 3.13, 3.14, and 3.15 branches.
  • The U.S. Treasury’s Financial Crimes Enforcement Network (FinCEN) has issued an order effectively cutting off MEXC Global‘s access to the U.S. financial system under Operation Economic Outcast. “MEXC Global is a convertible virtual currency exchanger that has facilitated at least $16 billion in illicit transactions connected to Iranian, North Korean, and Russian sanctions evasion, as well as cybercrime, including by transacting directly with sanctioned actors,” FinCEN said.
  • A new study of 25,964 U.S. government-related domains has found 43 whose DNS records referenced either mistyped, expired, or otherwise repurposed domains. In one case, a misspelled domain’s SPF configuration (“amzonses[.]com” instead of “amazonses[.]com”) authorized at least 32,066 unique IP addresses to send email on behalf of a U.S. county. “These external dependencies authorize third-party senders, route incoming mail, provide DNS delegation, and receive security telemetry,” SH Consulting said. “The government domains themselves can remain registered and be considered protected, because there are no signs to suggest anything is wrong.” SH Consulting said it defensively registered 17 of these dangling DNS dependencies to prevent abuse.
  • Abraham Teklu Lemma, 53, of Maryland and a former information technology contractor at the National Reconnaissance Office, pleaded guilty to conspiring to commit espionage on behalf of the government of the Federal Democratic Republic of Ethiopia. Lemma pleaded guilty to conspiracy to transmit national defense information to a foreign government from 2021 until his arrest in August 2023 and has agreed to a prison term between 120 and 262 months. “He exploited his IT administrator access to remove classified national defense information, including Top Secret information, from a secure facility and transmitted it to Ethiopian government officials hoping they would use it to target and kill paramilitary fighters in a regional conflict,” said Assistant Attorney General for National Security John A. Eisenberg.
  • Huntress has warned of a “couple waves of activity across multiple organizations involving GeoVision camera software,” pointing to broad exploitation of the platform to deliver malware. The activity originates from the IP addresses 45.32.117[.]34 and 156.246.94[.]183.
  • A single AI-augmented threat actor (AAT) has been observed using a mesh of 12 AI agents called BOSSAI to conduct persistent campaigns in several countries. The mesh consists of combining HexStrike AI, Hermes, CyberStrikeAI, and HalberdStrike, alongside Grok, OpenClaw, and Claude Code, with some acting as infrastructure builders and maintainers, while others scan for vulnerabilities and attack. The findings once again show how a lone actor can mount offensive operations at scale. “Using the BOSSAI mesh, the operator targeted hundreds of organizations spanning government emergency services, banks and lenders, universities, military systems, phone directories, e-commerce platforms, and gambling sites in eleven countries,” Acronis said. “It’s unclear if the attacker has any specific affiliation or goal beyond compromising as many organizations as possible, presumably to monetize the activity.” The threat actor is believed to be a single individual based in Hualien City, Taiwan.
  • Truffle Security has launched the Secrets Revocation Program, a free initiative that helps API key issuers revoke 1.4 million secrets that have been confirmed to be live and leaked on GitHub, Docker Hub, npm, Hugging Face, and others.
  • Gnosis Pay said it detected unauthorized access to the system powering its Gnosis Pay-branded cards on November 3, 2026, leading to the exposure of customer information. “The data breach was limited to the system powering Gnosis Pay-branded cards, which was already scheduled for retirement on 20 December 2026,” it said. “We have increased our monitoring of card activity and wallets to detect fraud or attempts to access user funds. We have also implemented an extra verification step (3D Secure) and made it mandatory for all online card payments. These measures will stay in place until the B2C system is fully retired as planned on 20 December 2026.” The company said it directly notified users about what information was accessed from their account, but did not reveal the number of customers impacted.
  • The U.S. Department of the Treasury’s Office of Foreign Assets Control (OFAC) sanctioned four individuals and 41 entities that are tied to scam compounds in Cambodia, Madagascar, and other countries across Southeast Asia and Africa. The sanctioned parties are said to have supported the business network of Prince Group. “Cyber scam syndicates continue to steal billions of dollars from American victims while relying on sophisticated international business networks to launder their illicit proceeds,” said Secretary of the Treasury Scott Bessent. “The newly designated individuals and entities have helped sustain Prince Group TCO’s illicit global operations after the October 2025 and June 2026 designations by managing offshore corporate structures, concealing beneficial ownership, facilitating commercial activities, and providing administrative and financial services to senior Prince Group TCO executives.”
  • The U.S. Treasury also sanctioned Chui Kok Wa for his role as the leader of a scam compound in Madagascar that was dismantled earlier this year. Late last month, Wa was charged with conspiracy to commit wire fraud, as the leader of the NFT investment scam. “In the scheme, scammers posed as beautiful women who met victims on dating websites, eventually enlisting them to connect their cryptocurrency wallets to fraudulent websites associated with the NFT investments,” the Justice Department said. “Once connected, the scammers willed the victims to continue to fund their cryptocurrency wallets, and then surreptitiously pulled money from the victims’ wallets without their knowledge.” Wa’s phones also allegedly contained employment contracts, detailing overnight work schedules in Madagascar and providing employees with bonus payments tied to money stolen, as well as scripts or guides for what scammers should say to victims and a roster containing employees’ names, passport information, salary, hours, disciplinary actions taken against them, along with a list of cryptocurrency wallet addresses for the employees.
  • Vercel has launched Security Review, a new capability in Vercel Agent that scans for vulnerabilities on every push to a team’s connected GitHub repositories. “Findings appear as comments on pull requests, are listed in the Security dashboard, and are available via CLI,” Vercel said. “With your permission, Vercel Agent can also validate and push fixes for findings.” In tandem, Vercel said it’s starting the rollout of Device Bound Session Credentials (DBSC) for vercel[.]com to make it harder for stealer malware to siphon session cookies from a device. Google rolled out DBSC support in Chrome earlier this April.
  • Fortra has revealed that attackers are creating convincing travel-booking pages that imitate trusted rental platforms. “The fake sites use familiar branding, realistic property listings, reservation details, and urgent messages to persuade travelers to provide sensitive data such as credit card information, usernames and passwords, and more,” Fortra said.
  •  Patrick Steven Yaroch, 37, has pleaded guilty today to wire fraud and money laundering related to using his access as an FBI agent to steal cryptocurrency. “From 2017 through early 2025, Patrick Steven Yaroch, 37, worked at the FBI Boston Division on an investigative squad focused on an adversarial nation-state,” the Justice Department said. “From February 2025 to July 31, 2026, Yaroch worked in the FBI’s Counterintelligence and Espionage Division, with access to FBI systems that contained cryptocurrency credentials that the FBI had collected as part of its investigations. Beginning no later than December 2024, Yaroch used his position to steal cryptocurrency from wallets controlled by the subjects of various other investigations or by adversarial nation-state actors, and from at least one wallet the FBI created to use in an operation. Throughout his wire fraud scheme, Yaroch conducted approximately 100 fraudulent transfers using credentials related to 35 separate FBI investigations. He illicitly gained approximately $2,474,380.66 in fraudulent proceeds.” The stolen funds were used to pay off his debts and finance a new life abroad, even asking ChatGPT how to maximize profit from investments and where to live in the European Union. Yaroch will be sentenced in March 2027.
  • DigiCert said it identified an event involving unauthorized access to one of its business systems, leading to exposure of names and email addresses. In some cases, driver’s licenses, passports, utility bills, and other documents containing a name and address that its customers provided as proof were also compromised. The incident has been traced back to a February 2025 change that moved the internal business system to a different network segment, causing it to be opened to the public internet for more than 16 months. Unknown attackers abused a self-registration feature in the system’s account manager to create a read-only account that had access to archived data between June 27 and July 2, 2026. Although DigiCert blocked the unauthorized account on July 2, it subsequently determined in October that the account was used to access archived customer contact information, including names, email addresses, phone numbers, and business addresses.
  • OpenAI announced 2 that it is increasing payouts in its Security Bug Bounty program. The objective is to get the broader research community hunting for flaws that its own testing may have missed. “A critical vulnerability in a core product such as ChatGPT, Codex, or the API Platform will now earn up to $30,000,” OpenAI said. “Exceptional critical findings on those targets remain eligible for up to $100,000, and now up to $172,000 with bonuses.”
  • Three people accused of using AI to impersonate corporate executives and attorneys to plunder about $129 million have been indicted on charges of wire fraud and aggravated identity theft. Jacob Soall, 41, Sharon Hazan, 54, and Guy Sheetrit, 42, targeted thousands of companies and attorneys at major law firms in order to trick employees into sending large payments to bank accounts under their control between 2024 and 2026, according to the Justice Department. Soall and Hazan were arrested in Spain in June and August 2026, respectively, while Sheetrit was arrested in the U.K. in September 2026.
  • A fake download campaign is targeting users searching for cracked commercial software, free games, activators, plugins, and unofficial utilities. The campaign uses trojanized Inno Setup-style installers hosted on bogus sites to distribute multiple payloads. “We confirmed three downstream monetization branches that transformed affected devices into monetized assets, including by routing third-party internet traffic through the device [WorldProxyPool], enrolling it in a distributed-computing platform [CharityEngine and BOINC], or installing additional utility-style software [AW Manager or Windows Manager], often without users’ informed knowledge or consent,” Microsoft said. “These products and components may have legitimate uses, but their silent installation through deceptive software bundles without informed user consent introduces security and privacy risks.”

  • A Chinese-language exploitation framework called Greeter has targeted Flowise honeypots with thousands of chat messages reading only “hello” while weaponizing an old code injection flaw (CVE-2024-58351) to execute malicious instructions. The framework conducted system discovery, extracted environment variables, probed cloud metadata services, searched for SSH keys and other credentials, tried to set up persistence, and ran quick checks to determine whether the host could run payloads from memory. Era Security said it counted 4,372 Flowise instances exposed online, out of which 703 (21.5%) had no authentication enabled. “It reveals a new exploitation framework designed to turn exposed AI applications into entry points for code execution, credential discovery, and potentially broader cloud compromise,” Era Security said. “The framework’s most consequential feature was its ephemeral, per-request execution model. Attack instructions were supplied through overrideConfig.javascriptFunction, while stored user messages contained only ‘hello.’ The malicious function could change with each request without altering the saved chatflow, leaving an incomplete picture for defenders inspecting workflow definitions or chat history alone.”
  • A threat actor has been abusing fake “suspended account” notices to hide phishing content from users revisiting its lookalike pages masquerading as business banking and treasury logins. The campaign, active since May 2026, has been relying on more than 300 fake domains, including 248 impersonating 189 U.S. financial institutions. The first domains were registered in late 2025, but most of them appeared between July and September 2026. “What ties them together is one mechanism: the kit remembers every visitor’s IP address, so the first visit gets a bank login page, and any return visit gets what looks like a closed hosting account. We call this cluster the ‘Suspended’ infrastructure,” KELA said. “Most domains follow the same recipe, [keyword]-[bank].com, with keywords such as secure, disable, protect and cancel, which makes the cluster findable in newly registered domain feeds.”
  • Meta said it paid nearly $1.2 million in bug bounties to researchers who found and responsibly disclosed issues in its Muse AI assistant. The bug bounty program went public in September 2026. “The largest single award was paid out four months before launch for a bug that allowed a researcher to get access to conversations and tokens in pre-release VMs used by employees and engineers testing the product,” Meta said. “An authentication check treated a header added by a local HTTPS proxy as proof that a request came from a trusted source. During a system migration, the environment gained a second proxy layer. That meant requests from outside could carry the trusted header too. A researcher found this and demonstrated the exploit.” Meta paid a $300,000 bounty for this report.
  • Chinese nation-state threat activity declined by 41.4% in the six months from April to September 2026, according to a new study by Trellix. The change is “a shift in operational focus, we assess, rather than reduced intent, and one that warrants vigilance rather than reassurance,” it added.
  • QNAP is urging users to update Plex Media Server on their NAS instances to the latest version available in the QNAP App Center following reports of active exploitation of a set of vulnerabilities known as Plex2Shell. “QNAP has received reports from some users that Plex Media Server on their NAS appears to have been compromised through these vulnerabilities,” it said. “The QNAP Technical Support team is assisting affected users and has confirmed that no data loss has occurred.”
  • MetaMask has disclosed that the security incident it identified on September 28, 2026, resulted in unauthorized access to part of its infrastructure, and that it took steps to contain the incident and secure affected systems. “We have identified no indication that MetaMask wallets or our wallet infrastructure were compromised as a result of this incident,” it said. “We have not identified evidence that the incident enabled unauthorized access to or withdrawal of customers’ staked principal. However, a limited amount of execution-layer staking rewards was diverted during the incident.” The attacker is said to have briefly altered the addresses used to receive execution-layer staking rewards for certain validators, diverting approximately 0.385 ETH in rewards. “These rewards are separate from customers’ staked ETH,” MetaMask added. “We have found no evidence that customers’ staked principal or withdrawal keys were compromised, and the affected systems have been secured.”
  • Infected security-themed GitHub projects are being used to target cybersecurity professionals to deliver backdoors and infostealer malware. According to Doctor Web, more than 2,000 PCs have been affected. The malicious code is embedded in Visual Studio project files, Python scripts, and Go source files. Once launched, the malware establishes persistence on the infected host, escalates privileges, disables security and recovery features, and deploys additional payloads, including AsyncRAT, a clipper malware, and an information stealer. The malware also attempts to remove ESET, McAfee, and Malwarebytes antivirus tools from the host and interferes with Windows backups. “The attackers behind these compromised development projects target blockchain game enthusiasts and (most intriguingly) information security professionals,” the Russian cybersecurity company said. “The first infected repositories emerged two years ago. The early uploads targeted crypto wallet users – back then attackers used repositories containing Visual Studio projects related to blockchain games, game bots and cheating.”
  • Security researcher Scott Helme has revealed that about 15,579 websites are currently infected with malicious JavaScript delivering credit card skimmers, ClickFix lures, iOS exploitation kits like Coruna and DarkSword, and malicious redirects. These websites receive more than 135 million visits every month.
  • A new report from Layer3 Intel has found that residential proxy traffic has exited through the White House network (AS2650) 236 times over a 30-day period between October 5 and November 3, 2026, with 28 separate residential proxy services able to route traffic through six of its addresses. “Most of the activity happens in the evening, Eastern time: about half of it falls between 5 p.m. and midnight, and almost none happens in the early morning,” Layer3 Intel said. “That pattern suggests a person rather than a public guest Wi-Fi network: someone regularly on the White House network late into the evening, whose device is infected with a proxy SDK. Our data shows the exits, not the device or the person behind them.” In a post shared on X, Synthient said “several White House staffers are running a Windows 11 machine with InfaticaSDK installed,” adding it found “a number of Windows 10 and Windows 11 devices beaconing to known proxy providers like Geonode, Infatica, and Mellowtel.” The root cause is believed to be Rave[.]io, a shared streaming app that bundles 12 proxy SDKs into the platform. Rave allows users to watch content from YouTube, Netflix, and Disney+ with friends and runs on Mac, iPhone, Android, and Windows.
  • A set of five vulnerabilities has been disclosed in Reachy Mini, an open-source, expressive companion robot from Pollen Robotics and Hugging Face, that, when chained together, can allow an attacker to “execute attacker code, escalate to a persistent root via the robot’s passwordless sudo, and ultimately gain control of the device’s motor,” JFrog said. “A second variant of the attack works through the browser, where the operator visits a malicious web page, which silently compromises the robot.”

  • A phishing campaign built around fake registration pages for OpenAI, SpaceX and Anthropic events is luring marketing employees with rewards like “3 months of ChatGPT Pro” or “3 months of Grok Heavy” to get them to sign in with their work accounts. “This campaign uses them in email-delivered invitations that lead to AitM and device code phishing, expanding the use of AI-themed lures across various payload types,” Push Security said. “Employees want access to the latest AI tools, making them excellent bait for attacks.” The Microsoft option launches a device code phishing flow run through the Kali365 phishing-as-a-service platform, while clicking “Continue with Google” opens a VoidProxy reverse proxy of Google’s sign-in page. It is unclear how the attackers initially distribute the phishing links. The kits used to build and deploy the fake event invite lure pages as ZoomMock and MSDRoot. Similar campaigns have been reported by ANY.RUN and Cofense in recent months.
  • A U.S. school estate was compromised by the 3AM ransomware group as part of a hybrid human-AI operation, according to Bitdefender. The exact initial access vector used is unknown, but the first recorded activity started on September 10, 2026. “Some parts show a human operator at the keyboard while others behave like an agentic loop, a language model reading each result and retrying with different commands,” Bitdefender said. “With the exception of the encryptor itself, the whole operation was built on legitimate, built-in tools, which is standard for ransomware operations. What is different is how the attacker behaved, and several things point to an agentic loop, a program that feeds each result back to a model and runs what comes back. The retries behave like a loop too.”

Conclusion

The strange part is how many of these attacks start with something small. A setting nobody checked, an update nobody questioned, or a system left online longer than it should have been. Add a little patience from the attackers, and suddenly a minor mistake becomes a much bigger problem. Not every attack needs a clever trick.

That’s it for this week’s roundup. Some of these stories will fade, while others may get worse as more details come out. Either way, there’s enough here to keep security teams busy. Check what applies to your systems, fix what you can, and we’ll be back next week with whatever turns up next.

Found this article interesting? Follow us on Google News, Twitter and LinkedIn to read more exclusive content we post.