Frozen snapshot — the feed as it stood week 46. Back to the live feed
Italy's data protection authority fined Meta €805 million after finding that Facebook for Android suggested photos and videos from users' galleries for posting without their explicit consent. The feature was switched on by default, and a single permission covered the whole app, so users were never clearly told their gallery was being accessed. Meta has 60 days to turn the feature off by default and ask for consent separately, and says it will appeal, calling the fine disproportionate. The Hacker News reporting.
Cursor confirmed attackers used a key exposed at its AI infrastructure provider Fireworks AI to steal internal data, then planted tampered copies of Anthropic's Claude Code extension in its marketplace on October 29. The malicious builds, versions 2.1.312 and 2.1.313, were live for about 11 hours and may have reached up to 5,000 developers, installing persistent SSH backdoors on their machines. Cursor warns that simply uninstalling the extension may not be enough, and has rebuilt its marketplace and moved its inference traffic off Fireworks. BleepingComputer reporting.
US authorities seized the cloud infrastructure behind FaceStealer, a North Korean social-engineering campaign that stole more than $400 million in cryptocurrency over 18 months. Attackers hijacked accounts on Telegram, LinkedIn, Slack and Discord to lure crypto and Web3 workers into fake video meetings, secretly recording their webcams and using deepfakes and ClickFix lures to plant malware. The FBI says it identified more than 900 secretly recorded webcam videos and 677 targets in 58 countries. BleepingComputer reporting.
CISA added two flaws to its Known Exploited Vulnerabilities catalog on Friday, both with a November 16 patch deadline for federal agencies. The UnrealIRCd flaw, CVE-2026-114713, lets any user on a linked IRC network inject commands that run with server authority — the project's maintainer calls it the most serious flaw in 15 years, and attacks began a day after the November 7 fix. The second, CVE-2026-123039, is the Check Point Gaia privilege-escalation flaw attackers used to reach root after breaking in through the PHP2Shell flaws. The Hacker News reporting.
Apple has published how Secure Exclaves — isolated runtimes in its silicon — keep camera and microphone data away from the main operating system, apps, and even Apple itself. The design powers face and voice recognition on the new HomeView display and the camera and microphone indicators on iPhone 18 Pro, iPhone Duo and MacBook Neo, with only narrow metadata crossing over to the OS. The exclaves run on a minimal microkernel Apple is formally verifying, a process that already caught flaws that could have led to full kernel compromise. SecurityWeek reporting.
PostgreSQL shipped updates fixing 52 security flaws, the worst a type-confusion bug, CVE-2026-6460, rated 8.8, that lets a database user who can create objects run code with the server's privileges. The bug's roots trace to 2004, and a researcher has shown a working exploit — though no attacks have been reported. The fix also requires rebuilding C-language extensions, and PostgreSQL 14 is now end-of-life, with 14.25 its final release. The Hacker News reporting.
Two attacks published this week break the promise of AMD's SEV-SNP confidential computing, letting a malicious cloud host read data inside VMs built to keep the operator out. Mior can also rewrite a VM's memory and fake the signed report that vouches for its software, while Skyfall leaks saved register values and even a disk-encryption key through processor timing. Both need control of the host's hypervisor, and AMD has firmware and microcode fixes dated August 12, with platform updates rolling out through December. The Hacker News reporting.
Trezor says three of its Safe 5 hardware wallets bought from reseller CryptoBilis contain a hidden extra circuit board that copies the recovery phrase from the wallet's screen and sends it over the mobile network to an attacker-controlled server. A firmware update, version 2.13.1, stops the implants found so far, but cannot protect a phrase captured before the update — anyone who bought from the reseller since March should move funds to a new wallet with a new phrase. No funds have been reported lost. The Hacker News reporting.
Shipping giant A.P. Moller-Maersk says an intruder got into a local legacy system inherited through an acquisition and extracted data, in an incident it disclosed on November 11. The company says the system is secured, the incident looks contained, and its findings so far strongly suggest no sensitive data was disclosed, with no impact on its wider systems or customer platforms. A ransomware group calling itself WhiteArmor claims it took about 100 GB and is demanding $5 million — a claim Maersk has not confirmed. The Register reporting.
An international police operation seized servers behind Gootloader, CastleRAT and StealC this week, taking down 235 servers and 257 domains and recovering about 6.4 million stolen logins belonging to more than 40,000 victims. Gootloader alone has infected at least 25,000 computers since 2018 and was used in US ransomware attacks that caused about $28.8 million in losses. No arrests were announced, and Dutch police have loaded the seized CastleRAT data into their Check Your Hack service so people can see if their email was in it. The Hacker News reporting.
Attackers are using the PHP2Shell flaws to break into unpatched Check Point Security Gateways, then escalating to root through a new Gaia OS privilege flaw, CVE-2026-123039, to plant a backdoor called TSGA. Check Point says the campaign began October 30 against Israeli organizations, widened worldwide on November 2, and in some cases ended with a wiper erasing files on VMware ESXi hosts. The company has not confirmed who is behind it, noting the evidence could also fit a false-flag operation. The Hacker News reporting.
Attackers began exploiting a new N-able N-central zero-day within hours of researcher Sean Bales publishing proof-of-concept code on November 12. The 4EverAble2Shell chain uses the Apache Axis flaw CVE-2026-90436, rated 9.8, to get unauthenticated remote code execution and then root on servers with the Legacy API enabled — including servers already on the latest hotfix. Researchers have counted 21 compromised servers mining Monero so far, and N-able has restricted the Legacy API while it investigates. SecurityWeek reporting.
Citrix patched 28 flaws in NetScaler ADC and Gateway on Thursday, including nine critical memory-overflow bugs that could crash appliances or make them behave unpredictably, plus a critical NetScaler Console flaw, CVE-2026-120869 rated 9.4, that lets unauthenticated attackers bypass controls and reach restricted resources. Fixed versions are 14.1-73.57 and 13.1-64.39 and later. Citrix reports no exploitation so far, but NetScaler flaws are a frequent target for attackers. SecurityWeek reporting.
Enea has detailed a previously unknown state-level actor, SilentAir, that hijacks nodes inside mobile operator networks to intercept targets' calls and text messages, mostly in Taiwan. The group routes a target's incoming calls and SMS through a compromised node and harvests subscriber details, using signaling attacks over SS7 and sometimes Diameter. Enea has watched the activity since early 2025, with hijacked nodes at operators in at least six Asian countries. SecurityWeek reporting.
Web-scraping company Apify confirmed that an attacker got into its internal systems and customer accounts on November 9–10, accessing personal information and stored data for some of its 91,000 customers. The attacker used a compromised credential to collect more credentials and move between accounts. Affected customers were told the exposed data included names, email addresses, salted password hashes, two-factor recovery codes and Google and GitHub sign-in tokens; Apify has not said how many accounts were hit. SecurityWeek reporting.
Attackers used cloud-hosted ConnectWise ScreenConnect panels to push rogue remote-access clients onto at least 288 endpoints across 48 organizations in about 90 minutes on November 11, Huntress reports. The rogue clients install a toolkit called SCKit that sets a trap to steal technician session cookies when they connect. ConnectWise disabled five third-party extensions across roughly 400 partner environments as a precaution; how the attackers first got panel access is still unclear. The Hacker News reporting.
A KrebsOnSecurity investigation has linked the administrator of Qilin — now the most active ransomware-as-a-service operation, with at least 365 claimed victims in three months — to a 36-year-old man in St. Petersburg, Russia. Researchers traced the handle "Haise" through old forum accounts, Telegram records and leaked Russian government-portal data to Aleksey Mikhailovich Yakovtsev. The identification is a research finding, not a charge, and no law enforcement agency has confirmed it. Cybernews reporting.
Public exploit code called SignDrop claims a standard Windows user can reach SYSTEM privileges by abusing the Common Log File System driver's new log-authentication feature. The flaw turns on how the driver creates a companion file when a privileged process — in the published code, Windows Defender — opens a prepared log, letting the attacker redirect that file creation. Microsoft has not confirmed the flaw, assigned a CVE or released a fix, and an independent test found the exploit works but not reliably enough for operational use. The Hacker News reporting.
Researchers at Island found 186 lookalike domains for AI tools — misspellings of ChatGPT, Claude, Gemini, Copilot and others — steering visitors into ClickFix lures, a browser-hijacking app called Net Guard, or search-hijacking extensions. One extension family has 16 live Chrome Web Store listings with 972,000 declared users. The pattern, dubbed TypoPrompting, shows how a single mistyped letter can now lead to a fake verification page that asks you to run a command. The Hacker News reporting.
Starting Tuesday, an attacker used stolen credentials for 4,837 GitHub accounts to plant a secret stealer in repositories those accounts could write to, according to Aikido Security. The stealer ran in at least 1,806 repositories, harvesting CI secrets and cloud credentials, and each attack began with a test pull request titled "docs: dependency audit notes" from a branch named wfx-docs. Affected owners should treat every secret their workflows can reach as compromised and rotate them — and check cloud accounts for backdoors the stealer plants to survive rotation. The Hacker News reporting.
A flaw in Apache HTTP Server's HTTP/2 module, tracked as CVE-2026-119618 and rated 9.8, lets a remote attacker run code on the server with no login — but only on servers with HTTP/2 enabled. The fix is in version 2.4.71, released November 11; Red Hat says public research has demonstrated a reliable exploit, though no attacks in the wild have been reported yet. Admins who can't update yet should remove h2 and h2c from the Protocols directive. The Hacker News reporting.
A flaw in MongoDB Server, CVE-2026-118383, lets an unauthenticated attacker send two requests that can crash a replica-set member if it later wins an election while still behind on replication — and turning on authentication doesn't stop the requests. Datadog researchers reproduced the crash in a lab; MongoDB rated it 9.5 and says it knows of no exploitation, with fixed versions out for every affected branch. Self-managed deployments need the fix on every replica-set member. The Hacker News reporting.
Node.js shipped fixes for 16 vulnerabilities this week, seven of them rated high, with the worst in its built-in HTTP/2 and HTTPS server code. One flaw lets unauthenticated clients crash HTTP/2 servers on the 24.x and 26.x lines, while another can make a server hand a client pieces of its memory — including responses meant for other connections and, in one HTTPS flaw, TLS private keys. Fixed versions are 22.23.5, 24.22.3 and 26.13.1, and end-of-life lines like 20.x get no fix. The Hacker News reporting.
Google Cloud says VMs running on AMD Turin chips need a guest OS patch and a stop-and-start to guard against three CPU side-channel flaws, including a new one, CVE-2026-0004, that may let code on a system infer data from the processor's L1 cache. Google has already applied mitigations between VMs and the host; the remaining risk is inside a VM, where an attacker with code on the same core could infer data from other processes or the kernel. Neither Google nor AMD reports any exploitation, and both rate the flaws medium severity with high attack complexity. The Hacker News reporting.
PTC has patched a critical flaw in its Windchill and FlexPLM software, CVE-2026-126325, rated 9.2, that could let an attacker run code through deserialization of untrusted data in crafted requests. The disclosure comes five months after a similar Windchill flaw was actively exploited and weaponized by Cl0p-linked actors in a data-extortion campaign — and researchers warn the software is a prime target for ransomware operators. PTC-hosted environments are already patched; self-hosted customers should apply the fix as soon as possible. The Hacker News reporting.
Zoom has patched two flaws in Zoom Workplace for Windows, including a critical improper-input-validation bug, CVE-2026-108186, rated 9.6, that could let an unauthenticated user escalate privileges over the network — though the victim must take some action first. The fix is in version 7.1.8, with a second flaw, CVE-2026-108187 rated 8.1, fixed in 7.1.5; Mac, iPhone and Android versions are not affected. Zoom reports no evidence of exploitation. The Hacker News reporting.
Researchers at Arctic Wolf say an AI-driven penetration-testing platform called AntiFox has been used in intrusions that stole AWS credentials and pivoted into cloud environments, in at least some cases reaching a production secret. The attacks chain web exploitation and web shells with automated credential validation and enumeration across dozens of AWS services, showing how agentic tools let attackers work at machine speed with little manual effort. The same family of tools has also been used to harvest Apple code-signing certificates and siphon cryptocurrency. The Hacker News reporting.
Microsoft's November Patch Tuesday fixed 1,259 vulnerabilities — the second month in a row over 1,000 — spanning 726 flaws in Windows, 174 in Office and 27 in SharePoint. The one to patch first is an Outlook flaw, CVE-2026-124194, that Microsoft says attackers are already exploiting and that can expose a user's NTLMv2 login hash. Microsoft has now fixed more than 5,100 CVEs this year. The Hacker News reporting.
A researcher calling himself forcequit published an iOS sandbox escape, worse_query, that lets an app read other apps' files on iOS 26 and 27 — including crypto wallet seed phrases and private keys. Apple's updates released Monday, iOS 27.2.1 and 26.7.3, do not list a fix, and Binance warned users to delete apps from untrusted sources and move funds to a wallet on a clean device if they suspect exposure. No attacks or thefts have been reported. The Hacker News reporting.
Apple released security-only updates across its lineup on Monday — iOS and iPadOS 27.2.1, macOS 27.2.1, Tahoe 26.7.3 and Sequoia 15.8.3, plus watchOS and visionOS 27.2.1 — fixing nine flaws on current iPhones. The worst is an iMessage bug, CVE-2026-113961, that could let a remote attacker join a group chat uninvited and read its messages; the updates also fix Bluetooth and AirPlay pairing flaws. Older devices get the same key fixes through iOS 26.7.3 and 18.7.12. TidBITS reporting.
Security firm Claroc found 37,571 apps built with Grok's chat builder online, and 35,578 of them were open to anyone without a sign-in — about one in ten held internal business or personal data. Among them were 24 apps built by staff at X and SpaceXSI, including a confidential sales onboarding plan and an ad proposal to a US federal agency; the company has taken those 24 down. The exposure comes from Grok's default "Anyone with the link" setting, and app addresses can be found through passive DNS records. The Hacker News reporting.
Platform provider Railway said AI provider keys stolen in an October breach were used to run up an estimated $740,000 in unauthorized spending, and it has promised to cover charges the AI providers won't refund. The attacker escaped build-system isolation on October 20–21, decrypted stored environment variables in up to 100,000 workspaces, and also took linked-account tokens for 1,852 users. Customers are being told to revoke and replace any sensitive variable stored on the platform before October 21. The Hacker News reporting.
Barracuda says attackers are now exploiting two flaws it patched on November 2 in its Email Security Gateway — an authorization bypass rated 9.3 (CVE-2026-116606) and a command-injection flaw rated 8.6 (CVE-2026-111560) — which chained together give an unauthenticated attacker root on the appliance. Public exploit code has been available since November 3, and CISA added both flaws to its exploited catalog on November 10 with a November 13 patch deadline for federal agencies. BleepingComputer reporting.
The Git project fixed a flaw, CVE-2026-111760, that lets a malicious repository run an attacker's code on a machine that clones it with its submodules — an over-long path slips past Git's safety check and gets written as ".git", letting the attacker plant malicious hooks Git then runs. The fix is in Git 2.56.1, with matching fixes for the 2.50–2.55 lines, and GitHub has already patched its own services. Until a fixed version is installed, GitHub advises cloning only from servers you trust and not recursively cloning repos you don't. The Hacker News reporting.
An attacker exploited a remote code execution flaw in SGLang, the open-source AI inference engine, and ran code on one of NEAR AI's inference servers — reaching requests processed there that day, including some sent in confidential mode, which is meant to keep them private even from the company. The attacker moved on to download account details for about 100 users and data from 18 AI agents on an older deployment. NEAR AI has revoked sessions and credentials, taken its Private Chat product offline and deployed a fix, but has not named the flaw's CVE or affected SGLang versions. The Hacker News reporting.
SAP's November Patch Day fixed four flaws it rates 10.0, the maximum score — two memory-corruption bugs in the NetWeaver web stack that let an unauthenticated attacker run code, a SAML flaw that could let an attacker sign in as any user, and a Gateway flaw letting an outsider pose as one of SAP's own servers. None of the four needs a login or any user action, and the researchers who found three of them say they have not seen them exploited yet — but SAP calls patching a priority. The Hacker News reporting.
Google pushed an emergency Chrome update after confirming attackers are already exploiting a use-after-free flaw in the browser's XML handling, tracked as CVE-2026-125350. The fix is in Chrome 157.0.8103.50/.51 for Windows, macOS and Linux, but Google rolls it out over days, so checking for the update manually is the fast path — and Google has not said who is being targeted or how the attacks work. BleepingComputer reporting.
CISA added four flaws to its exploited-vulnerabilities catalog on Monday, all with confirmed attacks in the wild: GPUStack's unauthenticated admin takeover (CVE-2026-102895), a FreeSWITCH code-injection flaw (CVE-2026-116779), the OpenSSH "SSHole" login bypass (CVE-2026-123158) and an IBM Aspera privilege-escalation flaw (CVE-2026-123218). Three of the four score 9.8, federal agencies must patch by November 12, and researchers have already recorded post-exploitation activity from the OpenSSH flaw, including stolen password hashes and a deployed command-and-control agent. The Hacker News reporting.
Independent researcher Sean Bales published details and working exploit code for 33 previously unknown flaws in N-able's N-central remote-monitoring platform, chained into four exploits — three reaching remote code execution and one escalating to root. N-able declared a critical incident, took its hosted N-central environments offline worldwide and told on-premises customers to shut their servers down until a hotfix ships. The company says it has seen no active exploitation yet, though researchers report exploit attempts against honeypots already. BleepingComputer reporting.
Vanderbilt University told students and staff to shut down university devices and disconnect from campus Wi-Fi on Monday while its IT team responds to an active cybersecurity incident. The ransomware group BYOD claims it locked the university's databases and stole more than a million Social Security numbers plus names, emails and student records — a claim Vanderbilt has not confirmed and no one has independently verified. Campus Wi-Fi was back by Monday evening, but most university systems stayed restricted. Neowin reporting.
Attackers broke into the AWS accounts of multiple Qubole customers between October 14 and 26, using credentials for an AWS account Qubole's data-lake platform uses to reach its customers — then stole stored secrets and copied data from S3 buckets. Wiz, which reported the breach, says even organizations that stopped using Qubole should check their CloudTrail logs, since roles left behind still trusted Qubole's account. How the attackers got Qubole's credentials has not been disclosed. The Hacker News reporting.
The N0n ransomware group listed Expedia Group on its leak site on November 9, claiming to have stolen loyalty-program tables holding 830 million rows — email addresses, member IDs, point balances and booking values — plus traveller profiles and password-like tokens. The claim has not been independently verified and Expedia has not confirmed a breach; what is confirmed is only that the listing exists, recorded by multiple threat trackers. Ransomware.live reporting.
LNbits, the open-source Bitcoin Lightning wallet system, told all operators to shut their servers down after a wave of attacks, saying it has several security fixes still to release. One operator reported an exploit withdrawing 2,285,545 satoshis before he could stop it, and the project switched off its own hosted service too; no fixed version had been released as of Monday morning, with patches expected within 24–48 hours. The Hacker News reporting.
Researchers Shubham Shah and Sam Curry found the login systems behind two car brands' apps ran on ForgeRock identity servers whose built-in admin console still used the documented default password, admin/admin. Reaching it through a path-parsing mismatch, they could take over any customer account and, on their own cars, unlock the doors and start the engine; both systems were fixed about two hours after they reported it. The Hacker News reporting.
Acronis researchers traced a single attacker in Taiwan who ran a 12-agent setup called BOSSAI against hundreds of organizations in at least 11 countries, pulling emergency dispatch records for seven Taipei districts and finding a fire-department platform exposing patient data since 2018. The agents ran mostly on locally hosted models with their safety training removed, after commercial models refused to help; they mostly succeeded where data was already poorly protected. The Hacker News reporting.
Frontline Education told Indiana regulators its August breach affected 3,085,266 people — more than twice the earlier state-by-state total — after attackers got in through a flaw in a third-party product. Stolen data from its Frontline Central and School Health Management products may include names, Social Security numbers, dates of birth, passport numbers and Medicaid IDs; the company has not named the exploited product or who was behind it. SecurityWeek reporting.
Amplifon Hearing Health Care disclosed that hackers broke into its internal systems in January 2026 and copied files between January 18 and February 8, affecting 1,332,826 people. The stolen data includes names, driver's licence numbers, dates of birth, hearing and medical information and health insurance details — plus Social Security numbers and financial information for some — though the company says it knows of no fraudulent use so far. SecurityWeek reporting.
Italian Foreign Minister Antonio Tajani said his personal data has been spread on the dark web for days, exposing him and his family to threats and intimidation, and Premier Giorgia Meloni blamed pro-Russian hackers. Tajani did not say what kind of information was shared, and the leak follows a string of hoax bomb threats and other incidents across Europe. SecurityWeek reporting.
Live feed holds the current week only — older stories roll off as each new week begins.