This newsletter is brought to you by Thinkst, the makers of the much-loved Thinkst Canary. You can subscribe to an audio version of this newsletter as a podcast by searching for “Risky Business” in your podcatcher or subscribing via this RSS feed. You can also add the Risky Business newsletter as a Preferred Source to your Google search results by going here.
For about 16 hours on Monday, November 16, the Optimizely A/B testing platform delivered malicious code on at least 178 popular websites across the globe.According to an Optimizely statement, the hacker exploited a vulnerability in its “Experimentation” platform to add custom code to certain customers.According to technical write-ups of what happened—via Report-URI, Step Security, and Amirul Iman—the code went through different versions across the day.It first delivered a clipper that replaced copy-pasted cryptocurrency addresses with ones controlled by the hackers in an attempt to hijack transactions; then added a module that collected payment card data entered online; then a module to collect browser cookies and screenshots of the active page; and later added versions of the Coruna and DarkSword exploit kits that tried to install malware on older iPhones.The malicious code reached a bunch of high-profile sites, like CNN, Zoom, Salesforce, GitLab, Equifax, HSBC, NPR, Fox News, Bitly, and the BBC. A full list is below, courtesy of Report-URI’s Scott Helme.According to security firm Hudson Rock, versions of the malicious script were also spotted on three other platforms, but they didn’t make it too far into their client list and were never loaded en masse, like they were from Optimizely. The three platforms are Kissmetrics, StorifyMe, and Minute Media.Optimizely now says it patched the bug exploited in the attack.At its core, Optimizely allows customers to load different versions of their sites and see which ones do better with their customers.This involves running a lot of JavaScript code from Optimizely that sets up those different versions and then measures how customers interact with each one.The bug was in a feature that allowed customers to load their own code on top of what Optimizely normal A/B testing feature was doing.Optimizely didn’t say exactly what the bug was, but security researchers have now confirmed that most of the infected sites were clean and safe to visit.Risky Business PodcastsIn this edition of Between Two Nerds, Tom Uren and The Grugq discuss why AI means that mercenary spyware is a dead industry. The pair discuss why governments would rather buy the cow than the milk.
Breaches, hacks, and security incidentsDutch sites warn of malicious code: At least 12 Dutch websites warned users that an advertising partner ran ads with malicious code. The code tried to collect identifying data from older iPhone models. Most of the sites are sports and recreation-related. [Autoreview.nl // NLTimes]Hackers breach Japan’s supercomputer: Hackers have gained access to two of Japan’s supercomputers. The National Research and Development Agency (Riken) says hackers stole information from its Fugaku and Rikyu supercomputers. Stolen data includes details for registered users, research data stolen in the home directory, and SSH keys for connecting the supercomputers to outside services. [Riken statement // NHK]Gemini tried to hack Epoch AI: Google’s Gemini cyber model attempted to hack the Epoch AI website at the end of October. The model performed the malicious actions in a test of its guardrails when it was told to retrieve secret AI benchmark results that were allegedly stored on the website. Google’s Gemini 4 Argon found and used canary tokens, attempted a SQL injection, and posed as staff to access the Epoch backend. Trajectory Labs, which conducted the test, apologized to Epoch AI and paused all misalignment testing. [Trajectory Labs]Fandango discloses breach: Ticketing platform Fandango has notified customers and authorities of an October security breach of its official website. [New Hampshire OAG, PDF]Coles beach: Australian retailer Coles has disclosed a security breach. Right now, it says no data was stolen. This could change in the future though. [Coles Group]Infoblox hacked: Hackers have breached and stolen customer data from cybersecurity firm Infoblox. Cybersecurity firm Infoblox is notifying customers of a security breach after hackers stole accessed data at one of its technology vendors. According to TechCrunch, the data was stolen from a company that provides SharePoint data migration services. The company didn’t say how many customers were affected or if it was being extorted. [TechCrunch]General tech and privacyCalifornia sues OpenAI over hacking spree: The California Attorney General has sued AI company OpenAI over a series of hacks this year. The lawsuit claims the company has put public safety at risk by failing to contain models in testing environments. California Attorney General Rob Bonta says his office is aware of at least 80 hacking incidents involving OpenAI. [California OAG]Microsoft mega-outage caused by July bug: According to a Microsoft exec, the recent cloud outage that caused M365 services to go down for almost an entire day was caused by a bug engineers introduced in the platform in a July update but which didn’t manifest until last Friday. [Charles Lamanna]Microsoft is testing an Edge Lockdown Mode: Microsoft is developing a new security feature for the Edge browser inspired by Apple’s Lockdown Mode. The new Edge Lockdown mode is an evolution of Edge’s existing Enhanced Security Mode. The existing mode disables the browser’s JIT JavaScript compiler to prevent common browser exploits. The new Lockdown Mode will also disable more browser processes that have been used in recent attacks. Microsoft will test disabling GPU-related APIs and later explore more. [Microsoft Browser Vulnerability Research]Windows Remote Lock: Microsoft has developed a new feature to let system administrators remotely lock a Windows device without wiping it. The new Remote Lock feature is currently being tested in Windows 11 Insider Builds. Once activated, the feature will sign out all the device’s sessions and lock the screen until an admin reactivates the device. [Windows]Government, politics, and policyFrance pushes for Iran sanctions over fridge hacks: The French government is pushing the European Union to sanction Iran over its refrigerator hacking campaign. More than 700 supermarkets across France were affected and lost tens of millions of euros in spoiled stock. French officials attributed the hacks to Iran’s Islamic Revolutionary Guard Corps Cyber-Electronic Command. In addition, Finland, Denmark, Estonia, France, and Lithuania have also called on fellow EU countries to sanction crypto networks helping Russia launder funds and finance its war in Ukraine. [Politico Europe // Finish Ministry of Foreign Affairs]EU sanctions NoName057(16): The European Union has sanctioned pro-Kremlin hacktivist group NoName057(16). The group emerged weeks after Russia’s invasion of Ukraine and spent most of its time launching DDoS attacks against countries that supported Ukraine’s defense. The same sanctions also target 24 individuals who worked for the Social Design Agency (SDA), a Moscow-based company running Russian influence operations. [EU]Germany passes new surveillance law: The German parliament has passed a bill to give its intelligence agencies extended hacking and surveillance powers. The new law allows agencies to hack foreign service providers, take down hostile servers, and carry out sabotage operations. The updated intelligence law also allows agencies to surveil foreigners and some categories of journalists and lawyers. The Bundestag passed the new law as a response to Russia’s invasion of Ukraine, its hybrid war in Europe, and after US criticism of outdated intelligence capabilities. [DW // Bundestag]US sets up AI alliance: The US and nine allies (Australia, India, Israel, Japan, South Korea, Qatar, Singapore, the UAE, and the UK) have signed a declaration to cooperate on AI technology and its supply chains. [US State Department]NSA network goes down: After major outages at Microsoft 365, the TSA, and several banks, it’s now the NSA’s Net network to be down. That’s the NSA’s internal comms network for handling classified communications and documents. The US is having a historical run of outages right now. [CNN]Russia threatens to block Steam: The Russian government has threatened to block Valve’s Steam gaming platform in Russia unless the company shares data on all its Russian users, including those who changed their country locale but still play from Russian IP addresses. [ProFinance.ru]Iraq and Jordan ban Telegram: The governments of Iraq and Jordan have banned instant messaging service Telegram. Jordan banned the app over the spread of extremist and unlawful content spread on the platform. Iraq banned Telegram for failing to remove reported content. [Al Bawaba // Jordan government]National Cyber Force HQ: King Charles III has opened the UK’s National Cyber Force (NCF) new headquarters in Samlesbury, Lancashire. [UK government]In this Risky Business sponsor interview, Catalin Cimpanu talks with Thinkst CTO Marco Slaviero about the state of the defensive AI cyber space and the company’s current agentic canaries, as well as the need for more experimentation.
Arrests, cybercrime, and threat intel8-Digits member extradited to US: The US government has indicted a Kazakh national for a ransomware and extortion scheme that targeted US schools, hospitals, and healthcare organizations. According to the Department of Justice, Roman Khlynovskiy was part of a group called the 8-Digits Team that worked as affiliates under the larger INC and Lynx ransomware operations. The group was involved in hacks that demanded more than $500 million in ransom demands. Khlynovskiy was arrested at a resort in Italy in July 2025 and extradited to the US this month. [DOJ // RiminiToday]GTA hacker pleads guilty: A member of the Lapsus$ hacking group has pleaded guilty to a number of online crimes, such as hacking Rockstar Games, BT/EE, Revolut, and Uber. Arion Kurtaj was initially sentenced to an indefinite stay in a secure hospital in December 2023 due to his severe autism diagnosis. He was released in January after doctors found him fit to face trial following a routine check last year. Kurtaj is widely known for releasing the source code of GTA5 and early footage from GTA6. [BBC]Judge clears MEV hacking brothers: A US judge has acquitted two brothers accused of stealing $25 million worth of cryptocurrency from crypto-trading bots. The acquittal comes a year after the judge ruled a mistrial after the jury couldn’t reach a verdict. Anton and James Peraire-Bueno were charged in 2024 for exploiting vulnerabilities in MEV trading bots to steal and launder crypto-assets. The two brothers claimed they only used their MIT education to come up with a novel trading strategy specific to the crypto medium and that they never hid their identities. [Reuters]India arrests pharma scammers: Indian authorities have arrested 12 suspects that operated a cyber scam call center. The suspects used hacked data from healthcare organizations to contact patients and sell counterfeit medicine. The group targeted patients with diabetes and tuberculosis across India. [The 420]Microsoft outage spam: Microsoft says spammers are exploiting its recent mega-outage to infect organizations with malware. [Microsoft]Wppconnect pushed malware: A threat actor has compromised the Wppconnect npm library for automating WhatsApp accounts and deployed malware that steals secrets and session data from a developer’s machine. [SafeDep]AI everywhere: A new Google report looks at how threat actors are exploiting AI for their operations. The report covers everything from LLMJacking to influence operations, supply chain attacks, and cyber espionage. It’s everywhere now. [Google]Japan finds 1k servers used for AI attacks: Japanese authorities say they’ve identified more than 1,000 servers that are being used to stage AI-powered cyberattacks against Japanese orgs. [Kyodo News]Tcherber profile: Cyderes researchers have published a profile on a cybercrime operation providing initial access services to hacked Fortinet devices, as well as a MaaS platform offering a custom C2 toolkit, a loader, and crypting services. [Cyderes]Malware technical reportsRetadup returns: Security firm Kaspersky has spotted a new version of the Retadup malware spreading in the wild. The new sighting comes seven years after French authorities seized the botnet in 2019. The old Retadup had infected more than 850,000 systems. The new one is currently on just 15,000 systems. [Kaspersky // Retadup takedown]TheMailer phishing kit: Researchers look at TheMailer, a self-hosted phishing platform that plugs into commercial AI models like Claude and ChatGPT at every stage of a campaign. The kit uses a feature that rewrites emails and retests delivery until a message evades spam filters and reaches the victim’s inbox. [Varonis]In this Soap Box edition of the podcast, Patrick Gray chats with Thinkst Canary founder Haroon Meer about his “decade of deception.”APTs, cyber-espionage, and info-opsMustang Panda in the Philippines: The Philippines CERT has warned government agencies that a Chinese hacking group named Mustang Panda is trying to hack government agencies ahead of the ASEAN summit in Manila. [NCERT-Philippines // Dream Security]TGR-STA-0049: A new cyber-espionage cluster has been targeting Southeast Asian government entities since at least July 2025. The campaign’s main tools are PerfectPioneer RAT and KUKURAT, unique RATs that use HTTP/3 over UDP as a C2 channel. [PAN Unit42]Rezet: Russian security firm F6 has published a report on the new TTPs used by the Rezet cyber-espionage group against Russia and other CIS countries. The group has been active since October 2018 and is also known as RareWerewolf, LibrarianGhouls, and LibrarianLikho. [F6]Operation ShadowCloud: Kaspersky has found a new cyber-espionage operation that uses cloud infrastructure to hide its operations. Per the company, this looks like a Chinese op, but no official attribution yet. [Kaspersky #1 // Kaspersky #2]
“Currently, we lack sufficient information to definitively attribute this activity to a specific threat group. However, we observed overlaps with activity from Chinese-speaking groups such as FamousSparrow and CopperTurtle (detailed reports on these groups are available to customers of our private reporting service). On several victim systems, these groups shared the same staging directories as the GitDoor backdoor, with their implants deployed within a one-day window of GitDoor deployment. While this does not prove a direct link, it suggests that a Chinese-speaking group may be developing and/or utilizing the GitDoor and GraphDoor backdoors.”
Pakistan adopts contagious interview tactics: A suspected Pakistani APT group is targeting Indians seeking employment at foreign defense contractors. According to security firm Lookout, the group uses fake job portals to lure targets into installing a remote access trojan on their smartphones. The malware is the GravityRAT, a common tool of Pakistani state hackers for more than a decade. [Lookout]Vulnerabilities, security research, and bug bountySecurity updates: Apache Shiro, Juniper, N-able, Oracle, PostgreSQL, Tails, Tor, uBlock Origin, Zyxel.NoisyTCP vulnerability: A new vulnerability can let threat actors crash internet-connected Linux servers using just a few malformed TCP packets. Named NoisyTCP, the vulnerability impacts Linux kernel TCP networking stack. NoisyTCP can be exploited without authentication and impacts all Linux versions released over the past 11 years. The Linux kernel team released patches last week. [GitHub // CVE-2026-99288]Apache Shiro patches major authentication bug: The Apache Shiro project has patched a vulnerability that allows users who logged in at least once to authenticate with any password. The vulnerability occurs when authentication caching is enabled for LDAP logins. Threat actors could abuse this bug (CVE-2026-129349) to log into Java apps if they knew a valid username. Apache Shiro is a Java security framework that handles authentication and authorization operations. [Apache Shiro]Tor patches bugs exploited for DDoS attacks: The Tor Project has released a security update to patch a bug that was being exploited to DDoS Tor servers. The patch fixes how Tor nodes handle IPv6 traffic. The bug was being exploited to flood Tor relays with fake client circuits and prevent real users from connecting. [Tor]Rootways exploitation: Hackers are exploiting a vulnerability in the Rootways Authorize.net CIM extension to take over Magento 2 online stores. [SanSec]uBO XSS: The uBlock Origin ad-blocker has released a security update to patch a critical universal cross-site scripting vulnerability. According to security researcher Smitty van Bodegom, a malicious website can use the bug to add malicious trusted filters to uBlock Origin extensions. The filter could then be used to run malicious JavaScript on any site the user is accessing. [uBO patch // Smitty van Bodegom]Mistral vulnerability: A vulnerability in Mistral’s Vibe coding assistant can run code on the users’ systems just by opening a malicious repository, all without user interaction. [Manifold Security]Anthropic changes disclosure timelines: AI company Anthropic is reducing its vulnerability disclosure deadline starting next year. The company will give organizations 30 days to patch high and critical-rated bugs and 45 days for everything else. The company currently allows companies 90 days to patch. The move comes after Google announced a similar disclosure timeline reduction starting next year. [Anthropic]Mior attack: A team of academics from Switzerland has unveiled a new attack that can break the confidentiality of AMD SEV-SNP trusted execution environment. The attack is part of a family of attacks known as the Interconnect Corruption Attacks (XCA) class, which also includes the likes of BreakFAST, Fabricked, and Staleus. [XCA Attacks]SKYFALL attack: Academics have developed a new micro-architectural attack against AMD CPUs that can be used to extract SEV-SNP guest state, recover decrypted microcode patches, and leak protected SMM states. [SKYFALL]Java4Shell vulnerability: Oracle has released security updates to patch a remote code execution security flaw in Java applications. Nicknamed Java4Shell (CVE-2026-114938), the vulnerability impacts Java SE 17 and later. Security firm Calif says it plans to release details as soon as the Oracle updates are out. This will leave very little time for companies to integrate the patches before attacks could potentially follow. [Oracle // Calif // Aikido Security // DataDog]
“CVE-2026-114938 is an out-of-bounds write in OpenJDK, meaning that any Java application running the affected framework versions may be exposed because the vulnerability happens before the application can reject any input.”
“1 PM Pacific today, after Oracle releases a patch.” So basically the moment Oracle releases a patch everybody’s Java is magically patched, right? Am I doing X wrong? :-)— Dr. Anton Chuvakin (@anton_chuvakin) November 17, 2026
Infosec industryThreat/trend reports: DNSFilter, Dragos, and SpyCloud have recently published reports and summaries covering various emerging threats and industry trends.New tool—KeenProxy: Ordo Systems has open-sourced KeenProxy, a tool for selective internet routing.
“KeenProxy lets you choose which devices, domains, IP ranges, and countries use a tunnel, connect directly, or are blocked. The web panel shows router load and lets you install, schedule, or roll back releases on managed Entware and native OpenWrt installations.”
OffensiveCon Tokyo 2026 videos: Talks from the OffensiveCon Tokyo 2026 security conference, which took place at the end of last month, are now available on YouTube.Risky Business podcastsIn this episode of Risky Business Features, Patrick Gray and The Grugq chat to Nazar Tymoshyk about Russia’s use of AI in recent state-back cyberattacks against Ukraine.
Click Here For The Original Source.
——————————————————–
