Ransomware gangs have been exploiting Palo Alto GlobalProtect VPN flaw since June
This week has already been rough for cybersecurity. Critical PHP and GitLab RCE flaws, serious Linux security bugs, compromised eMudhra certificates, a Node.js RCE flaw, a hacked LiteLLM release, and hundreds of hijacked Tor relays. That's a lot to deal with in just a few days. And the week isn't even over yet. But those big headlines are only half the story. Stolen passwords, exposed databases, broken login checks, and trusted tools being turned against their users are causing trouble too. Some attacks are clever. Others work because someone left a simple security gap open. This ThreatsDay, we're digging into the stories that didn't make as much noise but still deserve attention. Some are strange, some are serious, and a few might surprise you. Here's everything else you shouldn't miss this week. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out. 4,500+ accounts compromised Bluesky has acknowledged it's aware of a phishing attack that has compromised more than 4,500 accounts. "The attack spread through posts promoting a fake offer of a government payment," Bluesky said. "We have temporarily suspended the compromised accounts and are working with their owners to secure them and restore access. If your account was suspended, email [email protected] and we will help to get you back into your account." Android malware scam losses The Singapore Police Force has alerted users of a malware-enabled scam variant involving advertisements on social media platforms promoting senior-focused activities and programmes, water-based fitness activities, and recycling initiatives. The agency said at least 62 cases have been reported since April 2026, leading to S$989,000 in losses. "In this scam, victims would come across the fake advertisements on social media platforms, such as Facebook, and be instructed to download an Android Application Package (APK) file to facilitate programme registration or service arrangements," the Singapore Police Force said. "After installing the application, victims would receive unsolicited calls from scammers impersonating government officials or bank representatives, who would falsely claim that the victims' devices or bank accounts had been compromised, or that they were implicated in criminal activities. Victims would then be induced to transfer monies, disclose banking credentials or personal information, or surrender valuables for investigations." Customer credentials exposed Andromeda said it identified a security incident on October 13, 2026, that involved unauthorized access to credentials in its environment. "We disabled the compromised accounts that day and have observed no further unauthorized activity since," it said. "We believe the activity began as part of an industry-wide attack on or about October 9th. It affected a limited number of customer accounts and involved exposure of certain Andromeda and customer credentials. If we believe you were affected, you’ve already heard from us." The company said there is no evidence of unauthorized access to customer models, datasets, or training data. 20 security flaws patched Roundcube has released versions 1.6.20 and 1.7.5, addressing 20 issues including IMAP command injection, SSRF filter bypass, and unsafe attachment handling. There is no evidence any of the patched vulnerabilities have been exploited. Internal business systems breached On October 16, 2026, Genesys disclosed that it became aware of a cybersecurity incident involving an unauthorized third party that allowed them to access certain internal business systems, including a learning management system (LMS) and a customer relationship management (CRM) platform. The company said business and account information used to manage relationships with existing and prospective customers has been accessed and/or downloaded without authorization. However, it said its Genesys Cloud, Genesys Engage and Genesys Engage Cloud platforms were not compromised. Critical Kubernetes host takeover risk AegisSec Research has disclosed details of a NVIDIA AIStore for Linux flaw present in the legacy ETL Kubernetes PodSpec initialization API, CVE-2026-65162 aka ShadowForge (CVSS score: 10.0), that enables privileged container creation and host-level compromise. NFC bank fraud arrests Dutch police authorities have arrested two suspects in connection with a bank helpdesk fraud scheme that netted 170 victims and €380,000 in illicit proceeds. The arrested men include a 23-year-old from Vaassen and a 25-year-old from Ede. "Victims were called on their phones and played an audio clip," authorities said. "The scammers had created dozens of different audio clips using Artificial Intelligence, featuring various voices, banks, companies, and scenarios. The voice would, for example, alert them to suspicious activity on their bank account [and instruct them to] download an app and scan their bank card by holding it against their phone." In the process, the threat actors were able to forward the data from the bank card's NFC chip to the threat actor, who then used it to make unauthorized withdrawals. Critical API authentication bypass Laravel has patched a critical security flaw (CVE-2026-114509, CVSS score: 9.2) that could allow its built-in TokenGuard to be bypassed to authenticate as an arbitrary existing user without knowledge of their valid API token. "Under certain database configurations (MySQL/MariaDB), an attacker can submit a specially crafted, non-string credential value that is mishandled during the authentication lookup, causing the underlying query to match an unintended user record rather than failing authentication as expected," Laravel said. "An unauthenticated attacker may authenticate as another user without possessing that user's API token, leading to authentication bypass and account takeover on any route protected by the affected guard." The issue has been patched in versions 13.29.0 and 12.68.0. Unauthenticated database exposure An unauthenticated SQL injection vulnerability has been disclosed in GLPI 11.0.8 (GHSA-753w-2942-4pxm) that could allow a remote attacker to read the entire GLPI application database. "Depending on the version, this read access can lead to an administrator account takeover (detailed below) and remote code execution," Mobeta said. The vulnerability has been patched in version 11.0.9. Critical unauthenticated RCE risk Rapid7 said it discovered three new vulnerabilities in Eptura Archibus that could be exploited to achieve code execution. "A legacy API endpoint could be exploited to bypass authentication and run various functions without authentication (CVE-2026-107186, CVSS: 8.6 High)," it said. "The most serious functions allow for full database read (CVE-2026-107187, CVSS: 8.5 High) and an insecure file upload (CVE-2026-107188, CVSS: 9.6 Critical). The insecure file upload leads to remote code execution. A malicious actor with network access to the server via HTTP can read secrets and passwords out of the database. They can also run code to execute system commands as 'NT AUTHORITY/Local Service' and gain control of the server." 3.2 million secrets exposed Development platform Zeabur said it will shut down its services starting December 1, 2026, and finish by December 31, 2026, in response to a cybersecurity incident in which attackers leaked internal credentials stored in plaintext to compromise its infrastructure and steal over 3.2 million environment variable records and information for 321,810 accounts. "On October 14–15, 2026, an attacker used leaked Zeabur internal credentials to break into the shared cluster, some Dedicated Servers, and Zeabur internal systems, and eventually read the platform database," according to an announcement. This is the second time Zeabur has suffered a breach this year alone. In August 2026, the company said an attacker gained unauthorized access to its systems and stole sensitive data, including account information, API keys, and other secrets. Source code and customer data stolen PingCAP has disclosed that attackers used stolen credentials to access internal systems and copy source code and customer-related records, including SQL editor content and encrypted API-key values. The incident took place between September 9 and October 7, 2026. "An unauthorized party used stolen service account credentials to access certain PingCAP internal systems, copy data, and obtain source code," the company said. "PingCAP became aware of the incident on October 7 and began disabling the affected credentials the same day." Exported data includes user names and email addresses, organization and project associations, SQL editor text and errors, customer contacts, contract terms, orders, pricing, payment terms, and billing configuration. $2.5 million tax fraud scheme Chukwuemeka Victor Amachukwu (aka Chukwuemeka Victor Eletuo and So Kwan Leung), a 40-year-old Nigerian national, has been sentenced to eight years in prison in the U.S. for running a years-long hacking, tax fraud, and identity theft scheme. "Amachukwu took part in sophisticated schemes to hack into U.S. businesses, steal and use stolen identity information of victims, and commit fraud," the U.S. Department of Justice (DoJ) said. "Beginning in at least in or about 2019, Amachukwu, his co-conspirator Kinglsey Uchelue Utulu, and other Nigeria-based conspirators took part in a scheme to hack into U.S-based tax preparation businesses. The conspirators utilized spear-phishing emails to obtain access to these businesses' electronic systems. Once they had obtained access, the conspirators stole the tax and other identifying information for at least approximately 800 of the business's customers. The conspirators hacked into several U.S. based tax businesses, located in New York, Texas, and other states." The conspirators also obtained the stolen identity information of thousands of individuals and then used this information to file fraudulent tax returns with the Internal Revenue Service and state tax authorities, obtaining at least approximately $2.5 million. Military identity theft scheme Robert Wayne Boling, a 45-year-old U.S. citizen, pleaded guilty to running an identity theft and fraud scheme that defrauded thousands of U.S. service members and veterans while living in the Philippines. "Boling operated from the Philippines, and he received and exploited stolen personally identifiable information (PII) of U.S. servicemembers and veterans to compromise their bank accounts and steal money and veterans' benefits," the DoJ said. "Boling's co-conspirator, Fredrick Brown, who worked as a civilian medical records administrator for the U.S. Army in South Korea, took digital photographs of the PII of thousands of military-affiliated individuals. Brown provided the stolen PII to Boling, who, along with his co-conspirators, used the PII to compromise various accounts belonging to U.S. servicemembers and veterans, including bank accounts and online government-affiliated benefits accounts." Boling pleaded guilty to one count of conspiracy to commit wire fraud and one count of conspiracy to commit money laundering. He is scheduled to be sentenced on January 14, 2027, and faces a maximum penalty of 20 years in prison for each conspiracy charge. $6 million data breach settlement Chrome Holding Co., a holding company formerly known as 23andMe, has agreed to pay California $6 million to resolve accusations that it did not effectively secure customers' personal and genetic data, resulting in a massive 2023 breach that exposed sensitive information belonging to nearly 7 million people. "With regard to the genetic data or samples maintained by 23andMe Research Institute, consumers can request that the company delete their data, destroy their test sample, or revoke their permission for their genetic data to be used for research," the office of the California Attorney General said in a statement. Rogue AI containment guidelines U.S. agencies NSA, CISA, and NIST have released joint security guidelines to help organizations detect, contain, and shut down AI (or SI) agents that exhibit deviations from their approved tasks. The voluntary guidance sets out 10 practices for managing potential incidents during super intelligence training and evaluation, with a focus on maintaining human oversight and building public trust in secure SI development. "This voluntary document supports the frontier SI industry in preventing, preparing for, and responding to incidents during training and evaluation – particularly incidents in which one or more SI agents take significant uninstructed actions that compromise explicit or implicit policies, controls, or security measures imposed by the operator to define the intended task scope or to ensure human oversight," NIST said. Critical developer-machine RCE A critical remote code execution (RCE) flaw has been disclosed in OpenCode affecting Windows and Windows Subsystem for Linux (WSL). "The flaw allows an attacker-controlled remote Model Context Protocol (MCP) server to execute arbitrary code on a developer's machine during a remote MCP OAuth authentication process, without additional user interaction or consent," Pillar Security said. Anomaly, OpenCode's developer, fixed the issue in version 1.18.33. The vulnerability comes a few days after OpenCode confirmed a security breach affecting its Console, which allows teams to share access to AI models through centrally managed API credentials, potentially exposing API keys linked to users of Anthropic, OpenAI, and Google AI models. The incident, it said, was related to a "large-scale distillation attack that impacted 1.4% of OpenCode Console users." AI-generated code security bias AI models that have a pro-China leaning may systematically generate more vulnerable code when writing applications for Western organizations. "Ask for the app "for a US defense contractor" instead of "for a Chinese defense contractor", and Qwen3-Coder is 19% more likely to hand back an app that can be hacked," Hirundo said. "Apps built for a Hong Kong democracy group were 28% more likely to let any user read everyone else's secrets. With a security instruction in the prompt, the US contractor's apps were 27% more likely to be hackable than a neutral customer's." Hirundo also revealed that removing the model's CCP alignment removes the vulnerability from the model's weights. The development comes as Chinese open-weight models are being increasingly incorporated in enterprise workflows across the world. A similar testing by Booz Allen Hamilton in May 2026 found something similar: "Chinese LLMs produce more vulnerable code when prompted with a U.S. government persona than without — and the vulnerabilities are highly obfuscated Chinese LLMs inject PRC-aligned political bias into both the answers and code they generate." Hardware MFA downgrade phishing A phishing campaign is using Okta and ServiceNow-themed domains (riskified-okta[.]com and shipmonk-servicenow[.]com) hosting spoofed SSO portals and MFA downgrade lures targeting YubiKey users to target e-commerce SaaS providers. "The campaign pairs fake Okta login pages with an MFA downgrade lure targeting YubiKey users, allowing attackers to bypass hardware-based MFA where policy permits fallback methods like SMS or email codes," Dataminr said. "Targeting is concentrated on e-commerce infrastructure providers, likely because compromising these providers creates leverage for downstream third-party breaches and data extortion." Federal DeepSeek ban China's Foreign Ministry spokesperson Mao Ning said China "will resolutely safeguard the legitimate rights and interests of Chinese companies," adding "China consistently opposes the overstretching of the concept of national security." The development comes after the Trump administration banned DeepSeek-related technology from federal civilian agency IT systems, including third-party products and services built using its AI models. The DHS order, issued in July and made public on October 19, 2026, also bars federal civilian agencies from procuring DeepSeek products and services. According to the order: "DeepSeek servers are located in the PRC, and the PRC legal system permits extensive authority to government cyber and intelligence functions to access private company infrastructure. For DeepSeek-hosted instances of their models, DeepSeek's account generation process collects personally identifiable information (PII) and may also track and digitally fingerprint users. DeepSeek's most widely used applications function through a prompt-and-response process that requires presenting information to the model, and which may result in the transfer of sensitive data to the PRC if used by U.S. government employees or vendors. The transfer of the aforementioned information to the PRC, either independently or in combination with publicly available data, could produce actionable intelligence on U.S. persons and/or government programs. The models themselves may contain malware or other security vulnerabilities." XWiki exploitation and botnet activity Earlier this month, the U.S. Federal Bureau of Investigation (FBI) and Department of Justice (DoJ) announced the disruption of malicious tools used by a China-linked advanced persistent threat group known as Flax Typhoon. While Microscan was used to facilitate reconnaissance and computer vulnerability scanning, FishHub allegedly enabled the exploitation of computer networks through spear-phishing attacks and the deployment of follow-on payloads. In a follow-up alert, VulnCheck said it observed attempts against five vulnerabilities -- PHPUnit (CVE-2017-9841), XWiki (CVE-2025-24893), Apache HTTP Server (CVE-2021-41773), PHP-CGI (CVE-2024-4577), and ThinkPHP (CVE-2022-47945) -- from 119.13.79[.]145 (Hong Kong), 211.20.144[.]116 (Taiwan), and 210.242.76[.]32 (Taiwan). "The attacks originated from both Hong Kong and Taiwan," it said. "The Hong Kong IP exploited targets in Mexico, Great Britain, and Malaysia, while the Taiwan IPs exploited targets in India." The exploitation of XWiki by Taiwan IPs resulted in the deployment of reverse shells. On the other hand, the Hong Kong IP is a RedTail-family botnet. 600+ ransomware victims A new advisory from intelligence agencies in Germany, the U.K., and the U.S. has revealed that at least 600 organizations across North America, South America, Europe, and Australia have been breached by the SafePay ransomware gang since August 2024. Attacks involve the exploitation of known security flaws in edge device products and internet-exposed products (CVE-2020-3259, CVE-2023-20269, CVE-2023-27532, CVE-2024-40711, and CVE-2024-40766) or, alternatively, credential exposure and brute-force password attacks, for initial access, followed by persistence, privilege escalation, lateral movement, data exfiltration, and ransomware deployment. Back in August 2026, U.S. agencies revealed that Medusa developers and affiliates have impacted over 500 victims from a variety of critical infrastructure sectors as of April 2026, while continuing to exploit unpatched vulnerabilities as its primary means of initial access. The gang is also relying on access brokers, compensating them anywhere from $100 to $1 million to secure access to potential victims. However, there is no indication Medusa actors are developing their own zero-day or N-day vulnerabilities. HTTP connection warnings enabled Google has rolled out a change in Chrome where it "explicitly asks users for permission before making an insecure HTTP connection." This feature was enabled for users enrolled in Chrome's Enhanced Protection mode of Safe Browsing. It has now been extended to all users by default. "Chrome asking the user before making an HTTP request is the culmination of more than a decade of work from countless people and organizations,” it added. Critical call hang-up flaw A critical security flaw in the Signal messaging app (CVE-2026-113380, CVSS score: 9.3) can allow a malicious caller to keep audio and video tracks enabled even after the victim has hung up. The issue impacts RingRTC, a middleware library that provides Signal apps with video and voice calling services built on top of WebRTC. It also affects every Signal client across Android, iOS, and desktop. "A local hang-up is not a kill switch: it is just one more event placed at the back of the call's finite-state-machine event queue," a security researcher who goes by the online alias Imahian said. "The remote peer is trusted with that queue — there is no rate limit on peer-originated events and no priority lane for hang-up. Flood the queue with junk and the hang-up waits behind attacker noise forever, so the call (microphone, possibly camera) is never torn down." GitHub connector prompt injection A critical security flaw has been disclosed in OpenAI ChatGPT's GitHub Connector, which could trick the AI chatbot into using its authorized GitHub access to modify a different repository and potentially leak its secrets by means of an indirect prompt injection planted in an attacker-controlled public repository. When a bad actor prompts ChatGPT for a deployment review of their repository, it encounters the hidden instructions and proceeds to write a GitHub Actions workflow into another, private repository that has a ChatGPT connector with the default All-repositories grant. The workflow then accesses secrets and other sensitive information, and exfiltrates it to the attacker. According to security researcher Sunil Tripathy, the vulnerability stems from a combination of indirect prompt injection, an overprivileged ChatGPT connector, and the ability of GitHub Actions to automatically run arbitrary code when the agent makes a commit to the private repository. "It demonstrates cross-account write operations, CI execution and secret exfiltration via the ChatGPT GitHub connector specifically, using indirect prompt injection from a public repository with zero attacker access to the victim," Tripathy said. "The attack requires no command injection and no document upload — just a public repository and a victim who asks their agent to review it." Following responsible disclosure to OpenAI, the company described the issue as out-of-scope. The week isn't over, and neither are the threats. New bugs, fresh attacks, and unexpected security incidents can show up at any time. What looks safe today might be the next big headline tomorrow. That's why keeping up with security news matters more than ever. If you found this roundup useful, subscribe to The Hacker News, follow us on social media, and make THN part of your daily reading. We track the threats, explain what matters, and help you stay ahead of what's coming. Because in cybersecurity, the story you miss today could be the one that matters tomorrow.