Search This Blog
Powered by Blogger.

Blog Archive

Labels

Footer About

Footer About

Labels

Load More
Trendy Right Now

Popular Posts

Yoba
ChatGPT Claude cryptomining malware Yoba

Fake ChatGPT and Claude Apps Deliver Crypto-Mining Malware

 

Cybersecurity researchers have uncovered a malware campaign, known as Yoba, that uses fake desktop applications for ChatGPT and Claude to trick users into installing cryptocurrency-mining malware. The attackers created convincing GitHub projects and websites for supposed AI tools such as “GPT-5.6 Sol,” “GPT-6 Astra,” and “Claude Opus,” then distributed malicious ZIP files that appeared to be legitimate installers. According to Check Point Research, the campaign exploits public interest in AI tools and the trust people place in familiar technology brands. 

The malicious archives contain two key files: a genuine, digitally signed executable from OpenAI or Anthropic, and a harmful Windows DLL named winmm.dll. When a user runs the legitimate-looking Setup.exe, Windows loads the malicious DLL alongside it. This technique, called DLL sideloading, lets attackers abuse trusted signed software to make malicious code appear less suspicious and evade basic security checks. Importantly, the research does not suggest that OpenAI, Anthropic, ChatGPT, or Claude themselves were compromised. 

Once launched, the malicious DLL—called YobaDropper—checks that it is being run through the expected host executable, making it harder for researchers to activate it in a sandbox. It then copies a legitimate Windows media component, wmpdmc.exe, into a new folder and places another malicious DLL beside it. This second component, YobaLoader, establishes persistence through a scheduled task and retrieves further instructions from the attacker. 

YobaLoader uses an unusual command-and-control method to avoid detection. It downloads an image from a Telegram channel, takes the final 50 bytes as a seed, and generates a new domain using a domain-generation algorithm. It then queries Google’s DNS-over-HTTPS service for a hidden command stored in a DNS TXT record, which leads it to download the next-stage payload. By using DoH and common file-hosting services, the attackers make their traffic resemble ordinary internet activity and evade passive DNS monitoring. 

The downloaded payload is injected into Windows’ svchost.exe process, where it launches XMRig, a cryptocurrency miner. Victims may notice slower performance, higher CPU usage, increased electricity consumption, or unexpected system heat. Users should download AI applications only from official vendor websites or verified app stores, avoid installing “desktop versions” from random GitHub repositories, and be cautious of search results promoting unfamiliar AI tools. Organizations should also restrict installations from unknown sources and monitor endpoints for unusual scheduled tasks and mining activity.
Read more »
Government Data Theft
Cyber Attacks data attacks Data Breaches Data protection data security government data Government Data Breach Government Data Theft

Moldova Probes Possible Foreign State Involvement in Major Cyberattack Exposing Government Data

 

The Moldovan authorities are looking into whether the large cyber attack, which caused sensitive personal data to be breached and dozens of state services to be disrupted, could be linked to a larger operation targeting citizen data. Prime Minister Vasile Tofan said on Wednesday.

"We have seen coordinated attacks on a large scale. This shouldn’t be seen as an isolated incident. We’re looking at whether there was any foreign state actors involved," he said. The government has not named any countries or hacking groups. 

In total, hackers targeted vulnerabilities between October 19 and October 23 to gain large amounts of personal data from government services. They made around 11 million requests to access data. Preliminary findings show that information relating to around 750,000 identification numbers was removed but this does not necessarily mean all personal data was compromised. 

This could include personal information such as full names, date and place of birth, family relations, health insurance, social benefits, pensions, disabilities, drivers licences, images and car ownership. Financial information such as bank account numbers, tax liabilities and government payments could also have been accessed. However, authorities said there was no breach of bank cards and no evidence of the attackers attempting to tamper with official government documents. 

Initially the hackers exploited vulnerabilities within services that did not require authentication. This was then used to access more records from the government services that are linked. The government have reviewed 82 digital public services and have disconnected the vulnerable services. Over 30 government information systems and administrative services were also disconnected. Several public services were impacted including applying for building permits, veterinary transport authorisations, importing medicine and food safety registrations. 

Most of the services have now been secured, however they are being gradually reconnected following further security checks. On October 23 the authorities said unauthorised access was stopped. However, they said they identified further attempts on Sunday and Tuesday but these were unsuccessful. The government has warned that criminals could attempt to use the stolen information for fraud and impersonation. They could try to impersonate someone from a bank or public institution to get people to give them passwords or codes or send them money. 

The incident could be part of a larger hybrid attack campaign, government spokesperson Dumitru Ciorici said. Moldova has been constantly targeted by cyber attacks and disinformation since Russia launched a full invasion of Ukraine in 2022. 

Moldovan officials have previously accused Russia of launching disinformation campaigns to destabilise the country and weaken state institutions. They have also undermined confidence in the pro-European government led by President Maia Sandu. The latest incident has raised concerns over the security of Moldova’s digital services.
Read more »
Third-Party Security Risks
airline cybersecurity Amadeus Data Breach Aviation Cybersecurity Customer Information Leak Data protection Japan Airlines Data Breach Japan Cyber Attacks Passenger Data Exposure Third-Party Security Risks

Data Breach at Japan Airlines Adds to Japan's Mounting Cybersecurity Challenges


A data breach has been reported by Japan Airlines (JAL) involving approximately 92,000 customers following unauthorized access to its third-party technology provider, Amadeus IT Group, which exposed passenger reservation information. Due to several recent major cyber incidents reported by Japanese companies, this incident has raised further concerns about the security of customer data handled by external service providers. 

Japanese Airlines and Japan Transocean Air use Amadeus' passenger services systems for flight reservations, ticketing, and check-in operations. According to the airline, Amadeus informed them on October 23, 2026, that unauthorized access to their systems had occurred. This incident involved the breach of customer reservation records stored within the provider's environment. 

Passenger information included names, reservation codes, flight dates, routes, and travel classes of passengers that had been compromised. It is estimated that 90,000 customer information has been exposed. A second 2,000 passengers were also exposed with additional personal information contained in the compromised records, including passports, nationalities, frequent flyer numbers, email addresses, phone numbers, and flight, ticket, and seat numbers. 

According to reports, the unauthorized access occurred from August 10, 2023, to September 23, 2026, suggesting that the affected environment may have been exposed for an extended period of time. Although the method employed by the attackers to gain access has not been publicly disclosed, Japan Airlines has clarified that credit card information and passwords were not compromised in the incident. 

In addition, Amadeus has reportedly implemented security measures to address the issue, as the airline states there is no evidence that the exposed information has been misused and that flight operations have not been disrupted. According to JAL, affected passengers will be contacted individually and provided with necessary assistance. 

Customers were also warned against phishing emails, fraudulent phone calls and suspicious websites impersonating the JAL Group, since exposed travel and contact information might be used in targeted scams. In addition, this incident highlights the risk that third-party technology providers pose in the aviation industry. 

With its digital infrastructure providing airline, airport, hotel, and travel agency customers around the world, Amadeus makes security a critical concern for many organizations. Several airlines have reported possible exposures involving Amadeus systems, extending the impact beyond Japan Airlines. According to Korean Air and Asiana Airlines, unauthorized access to customer information managed by the same technology provider has been reported. 

It is reported that both airlines were notified of the incidents by Amadeus on October 20 and reported them on October 23. Based on their initial findings, it appears that reservation reference numbers belonging to some customers were exposed. Also pointed out was the unauthorized use of credentials related to an Amadeus system account. 

An incident similar to this one has also been reported by All Nippon Airways (ANA), another major Japanese airline. A notification of unauthorized access was received on October 29 by the airline and the provider is currently working to determine the extent of the exposure. As of now, ANA has not released the number of affected passengers or the specific categories of information affected. 

Once additional information has been confirmed, ANA will inform customers. These incidents have raised questions about the security of Amadeus's travel booking infrastructure as a result of their connection. As of October 1, the company disclosed a vulnerability in its travel booking display environment that may be exploited by ransomware actors. 

At that time, Amadeus stated that payment and other financial details had not been exploited by ransomware actors. Following the disclosures by multiple airlines, a closer examination of the earlier security issue has taken place. Amadeus provides technology services to more than 400 airlines, hotels, airport operators, and travel agents. 

In the event of a security vulnerability in a widely utilized platform, information could be exposed to several organizations, although it remains unclear how many airlines and passengers are affected. A person who has access to passenger names, travel schedules, and booking references may use these information to create convincing communications impersonating airlines using the exposed reservation information. This may also lead to opportunities for targeted phishing and identity fraud. 

It may be more credible for such attempts to possess additional information, such as passport numbers, email addresses, and telephone numbers. A broader cyber threat has prompted government intervention in Japan, as Prime Minister Sanae Takaichi called a cabinet meeting on October 26 to discuss the recent attacks and announced plans for emergency measures to strengthen cybersecurity protections for citizens and businesses. Her concern was expressed regarding the potential harm that might result from the misuse of personal information stolen. 

When Shinjiro Koizumi and U.S. Defense Secretary Pete Hegseth met in the Philippines on October 28, cyber security concerns of Japan were also discussed. Both sides agreed to intensify cooperation against malicious cyber activities that could adversely affect citizens, economic operations, and national security. 

Amadeus is still investigating these incidents, with the total number of passengers affected, the method by which the unauthorized access occurred, and the possibility of additional airline victims still being determined. It is evident from this incident that third-party technology providers in the aviation sector pose security risks. There remains an urgent need for stronger safeguards and a greater level of oversight in order to protect passengers' information as investigations continue to span multiple airlines.
Read more »
Ransomware
Data Breach energy company France GDPR Ransomware

French Energy Giant ENGIE Hit by Data Breach, Customer Records Accessed Through Partner Platform

 


ENGIE, one of the world's largest energy companies, has notified customers in France that an unauthorized third party accessed their personal data through a marketing platform used by the company's business partners, the company confirmed this week.

The breach notification letters, dated October 23, went to affected customers after ENGIE said it detected "unusual data access activity" on the partner tool. The company has not stated exactly when the intrusion took place or how long the attacker had access before it was caught.

According to the notifications, the data the intruder obtained includes customers' first and last names, mailing addresses, email addresses, and phone numbers. It also includes two types of energy infrastructure identifiers specific to the French grid: the Delivery Point identifier (PDL) and the Metering and Estimation Point identifier (PCE). Both codes are tied to a customer's physical connection to the gas and electricity supply network. Customer IDs and contract numbers were also taken.

ENGIE said no banking data was involved, and that the breach was confined to the partner-facing marketing platform rather than any of its internal systems. The scope of the incident is currently limited to its French customer base, though the company has not said how many customers received notifications.

"We have cut off the attacker's access to the platform, notified the relevant authorities, and are directly informing each affected customer," the company said in a statement.

Customers were advised to stay alert to unsolicited contact and potential phishing attempts in the wake of the exposure.

No ransomware group or extortion actor had publicly claimed responsibility for the breach at the time of publication.


A Company With a Recurring Target on Its Back

ENGIE operates across 30 countries, employs more than 90,000 people, and posted €71.9 billion in revenue last year. Its scale across utilities, renewables, and energy infrastructure makes it a high-value target for threat actors looking for customer data at volume.

This latest breach is not the first time attackers have made it inside ENGIE's wider corporate ecosystem. Approximately a year ago, the Scattered LAPSUS$ Hunters group claimed it had breached a Salesforce instance belonging to Engie Records, the retail energy subsidiary of ENGIE North America, and leaked records from the intrusion. Engie Records appeared on the group's leak site as the 27th of 39 organizations allegedly hit through the same Salesforce-related campaign, with around 3GB of data listed.

Before that, in August 2023, a separate incident affected roughly 110,000 French ENGIE customers. In that case, a hacker known online as "HommedeLombre" reportedly got into ENGIE's systems through an external contractor's network, a technique also used in the Pôle Emploi breach around the same period. That attacker exposed names, email addresses, and phone numbers, and later said the intrusion was meant to demonstrate that "nothing is ever secure," citing rising gas prices as a grievance. ENGIE filed a complaint with authorities following that incident and said it had brought the affected system under control.

If we look closer to understand the link here, the three incidents point to a recurring problem: ENGIE's data, whether held by internal systems, subsidiaries, or third-party partners, keeps ending up in attackers' hands.


Why Energy Sector Customer Data Is Valuable

The combination of contact information and energy-specific identifiers like PDL and PCE codes makes the stolen data more useful to attackers than a standard contact list. Those codes link a name and address to a specific gas or electricity metering point in the French national grid, which means the data could be used to convincingly impersonate ENGIE in phishing campaigns or to target customers with fraudulent offers around energy billing or contracts.

France's energy market also went through a turbulent period of price volatility in recent years, which means many customers have been actively comparing suppliers and receiving legitimate outreach about tariff changes. That context makes it easier for fraudsters to craft believable messages that reference contract numbers, delivery point details, and billing information, exactly the kind of data now in circulation.

ENGIE said it has notified the relevant authorities, as required under GDPR. French data protection rules mandate that breaches of personal data be reported to the CNIL, the national data protection authority, within 72 hours of detection, and that affected individuals be notified without undue delay when the breach poses a risk to their rights and freedoms.

As of now, ENGIE has not attributed the breach to any specific threat actor, and it is unclear whether the platform through which the data was accessed belonged to an external vendor or an internal partner system. The investigation appears to be ongoing.



Read more »
Healthcare Data
Cyber Networks Cyber Security Cyberattacks cybersecurity in healthcare Data protection data security Data Theft Healthcare Data

Cybersecurity Incident Disrupts New Mexico Hospital Network, Forcing Patient Diversions and Clinic Closures

 

A cybersecurity incident caused ambulance diversions, cancelled appointments and disruptions to medical facilities around the state on Monday October 26, leaving New Mexico with no emergency departments able to treat the most severe traumatic injuries. The University of New Mexico Health System provides the only level 1 trauma centre in the state. 

The health system advised patients not to attend the main hospital in Albuquerque and cancelled appointments at over 30 clinics whilst they tried to resolve the issue. Although officials did not immediately say what exactly had happened, later in the day it was announced that parts of the hospital and clinics were beginning to get back up and running. Presbyterian Healthcare Services and Lovelace Health System are also large healthcare providers in the state and worked with the network to prepare to take in extra patients. 

The University of New Mexico Health System sees around 900,000 outpatients and 100,000 emergency department patients per year (University of New Mexico Health System website), meaning this is a large part of the healthcare system. Multiple staff reported seeing messages on their computers late Monday morning stating that student, staff and patient data had been stolen. A spokesperson for the health system did not want to comment as they said there was an ongoing police investigation. 

An FBI spokesperson confirmed that they were assisting a partner organisation but did not give any further details. Attempts to comment were also sent to the Bernalillo County Sheriff’s Office and Albuquerque Police Department. This incident comes as there has been ongoing concerns about attacks on digital systems within essential services such as healthcare. Ransomware requests and other cyber incidents have been seen with hospitals for around a decade, however it is unusual for messages to be shown on computer systems due to the breach (Josephine Wolff, Professor of Cybersecurity Policy at Tufts University). 

This shows how difficult it can be for organisations to protect their networks. The impact could be seen throughout Albuquerque and surrounding areas. At the main hospital, staff in turquoise uniforms were directing patients away from the waiting area and advising they may have their appointments cancelled. Patients arriving at a medical centre in Rio Rancho were also told by security guards to rearrange their appointments. 

The University of New Mexico’s Health Sciences Centre, the largest academic medical centre in the state, cancelled Monday and Tuesday classes. Staff working in the hospital and medical group will be returning on Tuesday. It has not been confirmed what caused the incident, if data has been extracted or if there was a ransom request. The investigation is ongoing.
Read more »
Warp Breach
Agent Logs AI Security Data Breach Telemetry Exposure Warp Breach

Warp Breach Exposes AI Agent Logs and Telemetry

 

Warp has disclosed a security incident involving a third party infrastructure vendor that exposed AI agent decision logs and user telemetry. Chief executive Zach Lloyd estimated that approximately 1% of users could be affected, while emphasizing that the investigation remains ongoing. The disclosure identifies an exposure of operational records rather than establishing a compromise of every component in Warp’s development environment. The vendor, initial access mechanism, affected infrastructure components, and precise data boundaries have not been publicly identified, limiting independent assessment of the incident. 

Warp combines terminal functionality, a code editor, task orchestration, and agent activity review within an agentic development environment. Developers use its AI agents to generate code, execute tests, and deploy software. Consequently, agent traces and telemetry represent records associated with automated development workflows, although their exact contents vary with implementation. Warp reports at least 800,000 users, but its preliminary estimate of affected users does not establish the number of exposed records, compromised sessions, or affected organizations. Those measurements describe different dimensions of a security incident.

According to Lloyd, Warp detected what it believed was active, narrowly targeted exploitation of an infrastructure vendor on Friday and intervened within hours. The company described an attacker operation involving an “agent swarm,” sophisticated penetration techniques, and privilege escalation. However, the disclosure does not provide enough technical detail to reconstruct the attack chain or verify individual exploitation stages. No publicly identified vulnerability, authentication bypass, compromised identity, or execution mechanism has been specified. Claims about particular exploit classes would therefore exceed the available evidence. 

The exposed data categories were agent traces and telemetry. Warp has not established publicly whether these records included source code, user prompts, credentials, or other sensitive content. This uncertainty prevents a definitive assessment of confidentiality impact or potential secondary compromise. Exposure of operational logs should not automatically be equated with theft of API keys or unauthorized access to customer repositories. Warp notified identified affected users, informed the vendor, engaged CrowdStrike to assist with investigation, and contacted law enforcement. Further findings are needed to establish scope. 

A separate incident involving AI coding platform OpenCode illustrates why data classification matters when comparing breaches. That compromise exposed stored API keys for external providers, affected approximately 1.4% of users, and resulted in unauthorized charges reaching $1,700 for some victims. Equivalent credential exposure or financial abuse has not been established for Warp. The technical questions remain which records were accessible, what sensitive fields they contained, and how vendor access enabled their exposure. Until those details emerge, conclusions about downstream exploitation must remain provisional.
Read more »
Plausible Vulnerability
Brazilian Cybercrime Credit Card Theft Cyber Attacks e-commerce security Magecart Attack Online Payment Security Payment Card Skimming Plausible Vulnerability

Brazilian Magecart Group Exploits Plausible Installation to Steal Checkout


Data Magecart, a Brazilian group of hackers, compromised the analytics script used by e-commerce platform Convertr, potentially revealing payment information provided by customers to dozens of Brazilian fashion brands. During the attack, which was detected by cybersecurity firm Sansec on October 24, 2026, malicious code was injected into Convertr's Plausible analytics tracker. 

By modifying a JavaScript file loaded into Convertr's merchant checkouts, the attackers allowed the card skimmer to collect sensitive information during online purchases by modifying the JavaScript file. The data obtained was comprised of card numbers, expiration dates, CVV security codes, cardholder names, and Brazilian CPF taxpayer identification numbers. 

The information was transmitted to a server controlled by the attacker at 1x0.is. Online shopping and checkout services are provided by Convertr to dozens of Brazilian fashion retailers. According to the platform, its merchants have processed more than R$2 billion in sales. Due to the fact that the affected analytics tracker was distributed throughout its merchant network, compromising one file created a potential supply chain risk that extends beyond one individual online store. 

The malicious script was detected by Sansec after a clean version was captured earlier on October 24th. In October, the attackers modified the skimmer in order to target Convertr's payment API field names, making it more effective in capturing payment information. Researchers also reported that the compromised tracker continued to serve on October 26, despite efforts to notify Convertr. 

As Sansec has been tracking the Brazilian Magecart group since July 2024, the incident represents a significant development in its activities. The group has previously targeted individual online stores using Magento, OpenCart and WooCommerce, mainly in Brazil. With the Convertr compromise, Sansec has shifted from infecting individual stores to compromising a service shared by multiple businesses, resulting in the first supply chain attack attributed to the group. 

An unpatched installation of Plausible Community Edition, an open-source web analytics platform, is believed to have been the source of the attack. Versions 3.0.0 through 3.2.0 were affected by CVE-2026-8467, a critical vulnerability that could allow unauthenticated command execution through an exposed component of the /storybook component. 

Convertr reported installing version 3.0 in 2025 and not updating it prior to the incident. On May 15, 2026, Plausible released version 3.2.1 which fixed the vulnerability. Since researchers were not able to access the platform’s server logs, Sansec has not confirmed that the vulnerability was exploited to gain access. Therefore, it remains unclear what specific method was used to compromise the analytics server. Furthermore, the group has consistently targeted Brazilian payment systems as a focus for their methods. 

According to Sansec, its skimmers exhibit recurring characteristics, including the use of short-lived domains as recipients of stolen information, and endpoints containing Portuguese names related to card payments. Other variants have also been developed to attack payment services, including e.Rede, PagSeguro, Pagar.me and Mercado Pago. 

Another version is reported to replace genuine PIX payment QR codes with malicious codes. Convertr confirmed on October 27 that Plausible had been updated with a patched version following the discovery of the malicious script. In addition to notifying affected merchants, payment providers, and relevant authorities, the company also verified that the tracker hosted on Convertr's analytics server was free of infection. 

While it is important to remove malicious code, it does not necessarily eliminate all immediate risks for all shoppers. Sansec stated that browsers may retain cached copies of infected scripts for up to 24 hours following the server-side cleanup, potentially allowing the malicious code to remain active during that period. 

Payments made by Brazilian customers through checkout addresses containing checkout.convertr from October 24 onward should be blocked and replaced by their banks. It is also recommended to monitor transactions for suspicious activity involving CPF identification numbers in order to identify suspicious activity. 

A Sansec recommendation for organizations using Plausible Community Edition is to update the version to 3.2.1 or later and to review server access logs for requests to the exposed path of /storybook. Upon detecting evidence of exploitation, the affected server should be rebuilt, secrets rotated, tracker files checked against the official release, and a more thorough investigation conducted. The incident emphasizes the wider risks associated with shared web infrastructure. 

Multiple businesses can be affected by compromised analytics scripts without the attacker having to breach each merchant individually. To reduce the risk of similar attacks, it is vital that software be updated, scripts loaded on payment pages are monitored, and unauthorized outbound connections are restricted. 

This incident illustrates the dangers posed by compromised third-party scripts used on checkout pages. Preventing digital skimming attacks requires timely software updates, continuous script monitoring, and stronger protection of payment infrastructure.

Read more »
User security
Customer Data Cyber Attacks Data Leak Genesys User security

Genesys Hack Exposes Customer Data From Internal Systems

 

Genesys, a California-based provider of customer experience and call-center technology, has disclosed a cyberattack in which unauthorized actors accessed data from internal business systems. The company said the incident affected a learning management system and a customer relationship management environment, but stressed that its core customer-facing platforms were not compromised. 

According to Genesys, attackers likely downloaded business and account information used to manage relationships with current and prospective customers. The exposed information may include names, phone numbers, business email addresses, job titles, company and account details, and other customer-interaction data. While such records can be valuable for phishing and social-engineering schemes, Genesys has not said how many individuals or organizations were affected. 

The company said it acted quickly to contain the breach, including suspending the LMS platform. By October 23, Genesys announced that the incident had been fully contained, although its investigation into the scope and impact remains ongoing. It has promised further updates as it reviews the accessed information and confirms material facts. 

Importantly, Genesys stated that its Genesys Cloud, Genesys Engage, and Genesys Engage Cloud platforms were not affected. It said there was no unauthorized access to those platforms, customer and partner environments, or the data stored and processed within them. That distinction is significant because the company’s core platforms handle sensitive communications and operational data for enterprise clients. 

Genesys has not identified the attacker, but a newly emerged group called ScarlettGroup has claimed responsibility on its leak site. The company is urging customers to stay alert for suspicious messages impersonating Genesys and emphasized that it will never request sensitive information, such as passwords, through unsolicited email or other channels. The episode is a reminder that even when primary production platforms remain secure, internal systems holding customer-relationship data can still create meaningful privacy and fraud risks.
Read more »
money laundering
California cryptocurrency Cyber Fraud FBI money laundering

Chinese National Extradited to Face Charges in $36.9 Million Cryptocurrency Fraud

 


A 44-year-old Chinese national has been brought to the United States from the Netherlands to answer federal charges tied to a cryptocurrency investment scheme that drained nearly $37 million from 174 American victims.

Zhen Zhao was arraigned in California after his extradition, prosecutors said, following a federal indictment handed down in August 2025 charging him with conspiracy to commit money laundering. If convicted, he faces up to 20 years in federal prison.

Zhao is accused of running what investigators call a "pig butchering" operation, a type of confidence scam that has become one of the fastest-growing financial crimes in the world. The name reflects how the fraud works: operators spend weeks or months cultivating trust with their targets before stealing everything from them at once.


How the scheme worked

According to court documents, Zhao and his co-conspirators reached victims through unsolicited text messages, social media platforms, phone calls, and online dating services. The initial contact was typically low-stakes, a misdirected text, a friendly exchange that seemed accidental, or a casual introduction through a dating app. Over time, the operators steered conversations toward investment opportunities, eventually directing victims to what appeared to be functioning cryptocurrency trading platforms.

The platforms were fabricated. Built to resemble real exchanges, they showed account balances growing steadily over time, giving victims every reason to keep putting money in. When victims eventually tried to withdraw their funds, they were blocked, hit with invented fees, or simply cut off. Their money was already gone.

The ring funneled more than $36.9 million from U.S. bank accounts into an account at Deltec Bank in the Bahamas, registered under the name Axis Digital Limited. From there, the bank was directed to convert the deposited money into Tether, a dollar-pegged stablecoin widely used in fraud for its stability and ease of transfer. That Tether was then moved to a digital asset wallet controlled in Cambodia, where it was passed up to the leaders of scam centers operating in the region.

Cambodia has become a significant hub for this type of operation. Criminal syndicates there have turned large-scale cryptocurrency fraud into core business, often staffing compound facilities with trafficked workers from across Asia who are forced to run these schemes under threat of violence. The money chain in the Zhao case reflects that geography directly.


Eight accomplices, two already sentenced

Zhao was not running the operation alone. Eight co-conspirators have already pleaded guilty in connection with the same scheme. Two have been sentenced so far, one to 51 months and another to 46 months in federal prison. The remaining cases are moving through the courts.

The extradition itself was the result of months of legal coordination between U.S. prosecutors, the State Department, and Dutch authorities. The Netherlands, which maintains a functioning extradition treaty with the United States, processed Zhao's transfer back to face trial in California.


A problem that keeps growing

This case sits inside a much larger crisis. The FBI's 2025 Internet Crime Report recorded more than $8.6 billion in losses from investment scams over the course of a single year, up from $6.5 billion the year before. Those figures came from roughly 73,000 complaints, and investigators routinely note that reported numbers reflect only a portion of real losses, since many victims never file a report out of embarrassment or because they do not know where to turn.

Pig butchering has become the dominant form of cryptocurrency fraud in the United States. The FBI ran a proactive victim notification program that reached more than 4,300 pig butchering victims in a single year, warning people before they lost everything. Despite those efforts, the dollar figures keep climbing.

The Department of Justice has treated these cases as a priority. In December 2023, federal prosecutors in California charged four people with laundering more than $80 million connected to similar cryptocurrency investment scams, using shell companies and hundreds of individual banking transactions to move stolen money internationally. Prosecutors at the time described the mechanics as nearly identical to what Zhao and his network allegedly ran.


What victims should do

Anyone who believes they were targeted by a pig butchering scam can file a report with the FBI's Internet Crime Complaint Center at IC3.gov. Investigators ask that victims preserve all records of communications, platform names, cryptocurrency wallet addresses, transaction IDs, and any banking information they shared during the fraud.

For Zhao, the arraignment in California is the start of what will likely be a long legal process. With eight alleged co-conspirators already pleading guilty, prosecutors appear to have a functioning case built around his network. Whether he opts for trial or takes a similar path remains to be determined.




Read more »
Web Hosting
Account Compromise Cyber Security Data protection DreamHost Password Reset DreamHost Security Incident email security FTP Security Hosting Security Web Hosting

DreamHost Forces Password Resets Amid Ongoing Security Incident


Due to a disruption of access to DreamHost's File Manager and webmail services as a result of an incident, certain email and FTP/SFTP accounts were forced to reset their passwords. Although the hosting provider described the action as a precautionary security measure, it has not yet revealed the cause of the incident or confirmed whether any accounts have been accessed without authorization. 

When DreamHost first detected connectivity problems affecting its file manager on the web at files.dreamhost.com on October 23, 2026, the disruption was initially reported. As a consequence, some webmail users experienced login failures, while FTP/SFTP access was also affected by the issue. As a result, the company reset passwords for affected accounts and temporarily restricted access until new credentials could be established. 

When a DreamHost account owner assigns a replacement password to the affected mailbox and file transfer users, the users cannot sign in. By the weekend, DreamHost mailed recovery instructions to all affected account holders in stages and confirmed to all affected account holders that all were notified by October 25. 

In October, the company reported that engineers had deployed a fix and had restored File Manager functionality. Monitoring continued to ensure that the service continued to operate, however password changes for affected accounts continued to be required. Since limited information has been released by the company about the incident, customers have expressed concern. 

Neither DreamHost nor the company has identified the root cause, disclosed the number of accounts affected, or confirmed whether any email messages, stored files, or login credentials have been compromised. Also, the provider advised users to review their mailbox forwarding settings and remove any unfamiliar destinations. This setting can be abused in order to redirect sensitive communications outside a company, thereby making it a crucial element of account security. 

DreamHost has, however, not confirmed that unauthorized forwarding rules have been created in this incident. Additionally, customers have expressed concern about the lack of detailed explanations for the disruptions in service as well as mandatory password changes. Despite speculation about a possible compromise, no publicly available confirmation has been provided to indicate that an attack or data breach has occurred. 

According to DreamHost, account owners should update stored credentials across connected applications and review account settings in order to complete the recovery process. Email clients, FTP/SFTP software, website backup tools, deployment systems and plugins relying on previously stored credentials may be affected by password changes. 

In order to avoid the possibility of affected users losing access to their hosted mailboxes, the company recommends that replacement passwords be shared through private channels, such as telephone calls or secure text messages. The recovery guidance provided by DreamHost specifies that mailbox passwords can be changed under “Mail > Manage Email” in the control panel, and FTP/SFTP credentials can be updated under “Websites > SFTP Users & Files”. 

If you changed your password before the reset process was completed, you should verify that your new password remains valid as well. Password changes may take up to 15 minutes to become effective. There is no indication that incoming email is reaching affected mailboxes, and that FTP/SFTP data is still stored on the provider's servers. 

In addition to these assurances concerning service and data availability, it is also recommended that website owners conduct routine site checks to determine whether unauthorized access occurred prior to the password reset. Considering that FTP/SFTP credentials provide access to website files, it is important to identify suspicious changes by reviewing recently modified files, unfamiliar administrator accounts, and unforeseen scheduled tasks.

There has been no evidence published to indicate that DreamHost-hosted websites were altered as a result of the incident. This episode also underscores the importance of using third-party hosting providers for security reasons. When a hosting company encounters a disruption, several connected services may be affected at once, especially when email addresses and file transfer credentials are handled by the same provider.

In order to ensure that essential operations are resumed after the password change, DreamHost has repeatedly warned customers that it will never send customers passwords by email. Businesses using DreamHost may need to check integrations and automated processes reliant on these accounts. Rather than logging into the control panel through unsolicited messages, account owners should access the control panel directly. 

No public explanation has yet been provided by the company as to what triggered the incident or the number of affected customers has been confirmed. In the interim, a central question remains as to whether the incident involved unauthorized access, how many accounts were affected, and if any customer information was revealed. Despite DreamHost's password reset process and restoration of File Manager access, the broader issues associated with the incident haven't been resolved. 

DreamHost's restoration of File Manager access marks progress in the recovery process, but the absence of an identified root cause leaves the incident's broader security implications unresolved. While the company has yet to clarify the cause of the incident, and whether any customer accounts or data has been accessed, it remains unclear how the incident will impact the security of the company.

Read more »
GPT-4
AI Model AI Security AI technology ChatGPT. OpenAI Cyber flaws Cyber Security GPT GPT-4

Researchers Uncover Compaction Smuggling Flaw That Exposes OpenAI Reasoning Traces

 


Researchers at Xint identified a vulnerability in the OpenAI API that could expose hidden reasoning traces from advanced AI models, creating a potential route for large-scale model distillation. The technique, named “compaction smuggling,” bypassed safeguards introduced after earlier research demonstrated methods for extracting encrypted reasoning data. OpenAI confirmed the issue and deployed a fix on October 23, 2026. Xint received a $6,000 bounty for the discovery. 

Reasoning models such as GPT-5.6 Sol generate intermediate reasoning before producing answers. OpenAI does not provide API users with the complete raw reasoning in readable form, instead offering optional summaries. This restriction helps protect against distillation, a process in which developers train smaller models using outputs from more advanced systems. Full reasoning traces can provide particularly useful training data because they contain intermediate steps, incorrect approaches and self-corrections. 

The API stores raw reasoning in encrypted items that clients can return during later interactions but cannot directly decode. OpenAI also encrypts conversation compactions, which condense lengthy exchanges into smaller items while preserving relevant context. The discovery follows an August 2026 study titled Stealing Reasoning Traces from Proprietary LLM APIs. The researchers behind that study demonstrated that encrypted reasoning could be replayed into weaker models from the same provider and reproduced as text. 

The approach affected models from OpenAI, Anthropic and Google, although OpenAI’s safeguards made extraction more difficult. Subsequent protections reportedly stopped the original technique from working. Xint discovered that the compaction feature offered another route. The researchers first generated reasoning using GPT-5.6 Sol and retained its encrypted reasoning item. They then removed the original task from the conversation history and instructed the compaction process to preserve the reasoning verbatim in a JSON object. Replacing spaces with tab characters helped the content evade usage-policy classifiers. 

The resulting encrypted compaction was then submitted to GPT-4.1 and GPT-4o, which were instructed to reproduce the contents exactly. After decoding the JSON and converting tabs back into spaces, the researchers recovered the reasoning trace. The technique worked across all tested reasoning-effort settings, from minimal to maximum. In the demonstration, the original reasoning item contained 405 tokens, compared with 612 tokens in the compaction and 609 tokens in each echoed output. Xint compared outputs from two models using different prompts and obtained identical results. 

Although the original plaintext was unavailable for a definitive token-by-token comparison, the recovered trace retained its false starts, corrections and characteristic writing style. The researchers developed the technique while investigating GhostCompact and examining how encrypted compactions could be transferred between models. The findings raise concerns about large-scale extraction of proprietary AI capabilities. 

In September 2026, U.S. agencies warned about industrial-scale model distillation campaigns. Anthropic subsequently reported seven campaigns attributed to China-based laboratories, while OpenAI disclosed disrupting an effort that generated approximately 16,000 requests from more than 4,000 accounts in July. 

Xint concluded that server-side processes handling decrypted reasoning while accepting client-controlled instructions can become potential extraction pathways. The researchers said compaction features require protections comparable to those applied to raw reasoning traces.
Read more »
NFC Relay Attack
Bank fraud Contactless Payments Cyber Crime Dutch Police NFC Relay Attack

Dutch Police Arrest Two Over $400,000 NFC Relay Bank Fraud

 

Dutch police have arrested two men suspected of operating a sophisticated bank-fraud scheme that used NFC relay attacks to steal more than €380,000, approximately $425,600, from 170 victims. The suspects allegedly combined AI-generated robocalls, fake bank employees, and malicious mobile applications to trick people into exposing their contactless debit-card data. The arrests highlight how criminals are adapting near-field communication technology, which is designed for convenient tap-to-pay transactions, into a tool for remote account theft. 

According to Dutch authorities, the scam began with automated calls that used AI-generated audio to impersonate banks, companies, or other trusted organizations. Victims were told that suspicious activity had been detected on their accounts and were asked to press “1” if they were not responsible. Once the victim responded, the call was transferred to a suspect posing as a bank employee, who claimed that the person’s debit card had been blocked and needed verification. 

The fraudsters then instructed victims to download an application and hold their debit card against their phone. Rather than simply verifying the card, the app extracted data from the card’s NFC chip, which supports contactless payments. That information was relayed in real time to an accomplice waiting at an ATM, who used it to withdraw money from the victim’s bank account. This technique is known as an NFC relay attack because it effectively extends the card’s contactless signal across a distance. 

Police arrested a 23-year-old man from Vaassen and a 25-year-old man from Ede on October 6. Investigators believe the younger suspect likely built the technical infrastructure behind the operation and orchestrated the attacks, while the second suspect played a coordinating role. Authorities also took the criminals’ servers offline and seized them, along with phones and laptops believed to be connected to the scheme. Both men remain in custody while the investigation continues. 

The case is a warning that a phone call claiming to be from a bank should never be treated as proof of identity. Genuine banks generally do not ask customers to install unknown apps, scan their cards against a phone, or share one-time codes. Consumers can reduce their risk by hanging up and calling their bank using the number on the official website or card, avoiding app downloads from unsolicited calls, and treating any request to tap a payment card on a phone as a major red flag.
Read more »
OpenAI
ai agents Anthropic Claude Code malware OpenAI

AI Coding Agents Can Be Hijacked by a Single Malicious File in Your Downloads Folder

 



A security firm has found that placing a disguised instruction file inside a user's Downloads folder is enough to manipulate AI coding agents into installing malware, with one leading tool executing a known-malicious package in more than nine out of ten test runs.

The finding comes from Forever Security, which said one of its customers had an `AGENTS.md` file sitting in their Downloads folder. The file was designed to look like routine engineering notes, but its actual purpose was to instruct any coding agent that opened the folder to install `function-color`, a package tied to the MALFEX npm supply chain campaign. Forever stopped the agent before the installation completed. The firm could not determine how the file got there.


How the Malware Chain Works

`function-color` is a wrapper package. It carries no payload of its own but silently installs `function-flag` as a dependency. That package runs a postinstall script that downloads a Windows executable from a remote host, saves it as `node.exe` inside `%APPDATA%`, and executes it in the background. The download routine sits inside an empty catch block, so a failed attempt produces no error and leaves no trace in the install log. As of early October, both packages remained installable on npm with no advisory flagging either one.

The reason the attack targets the Downloads folder specifically is not arbitrary. Browsers can save files to that directory without asking the user for permission or even requiring a click. Forever built a proof-of-concept web page that silently dropped a crafted `AGENTS.md` into Downloads during a normal browsing session. The firm did not establish that an attacker used this exact method against its customer, but the capability is real and requires no special access.


Test Results Across Two Major Agents

To measure how reliably the technique works, Forever ran controlled tests against OpenAI's Codex and Anthropic's Claude Code.

For Codex, the planted file instructed the agent to register a new server entry in `~/.codex/config.toml` and install the malicious package. Because that config file applies to every session on the machine, a successful write would persist across projects. Codex followed through in more than 90 percent of the test attempts inside an isolated virtual machine.

Claude Code required a different approach. The planted `CLAUDE.md` was written to resemble internal gateway setup documentation, and its goal was to write a custom API endpoint into the project's `.claude/settings.json` file. If an attacker controls that endpoint, every API call from the agent passes through their infrastructure, along with the session token. The token can then be used to take over the account. Claude Code normally prompts the user to approve writes of that kind, and it frequently rejected requests that looked suspicious. The extended engineering-document framing, however, got the write approved roughly half the time.


Neither Vendor Has a Patch

Forever stated plainly that there is nothing for OpenAI or Anthropic to fix. Each agent was doing what it was built to do: read instruction files in the working directory and carry out the steps listed in them. The firm compared the situation to a browser rendering a phishing page. The browser is not broken; the page is malicious.

That framing shifts the responsibility to users and organizations. Forever's own telemetry adds urgency to that point. About half of its customers run a coding agent from their Downloads folder, and roughly a quarter of agent sessions involve an open browser. In many of those cases, no person is actively watching the session when a new file arrives in the directory.


What to Do

The most direct fix is to stop running coding agents from the Downloads folder entirely. Beyond that, Forever recommends reading any `AGENTS.md` or `CLAUDE.md` file you did not write yourself before running a session, including files that arrive in freshly cloned repositories.

For anyone who suspects a file may have already been processed, the firm advises checking `~/.codex/config.toml` for unrecognized server blocks and reviewing `.claude/settings.json` for API endpoints that were not intentionally set. On Windows machines, the presence of `node.exe` inside `%APPDATA%` is a strong indicator that the function-flag payload ran successfully. If a planted file is confirmed, the session's login token should be treated as compromised and rotated immediately.

Keeping approval prompts enabled and running agents inside sandboxes reduces the attack surface, though the Claude Code test results show that approval prompts alone are not a reliable barrier when the malicious instruction is written to look like legitimate documentation.



Read more »
Malicious Apps
2FA bypass Account security Andriod Android Spyware Android spyware protection Cloning Cyber Security Login Credentials Login Security Malicious Apps

uCreator Android Spyware Targets Telegram and WeChat Accounts With Credential Theft and Cloning

 

A commercial Android surveillance toolkit known as uCreator can compromise Telegram accounts by capturing two-factor authentication (2FA) passwords and duplicate WeChat accounts through malicious app modifications, according to security researcher Ovi. The toolkit reaches users through tampered applications distributed outside official app stores or through a container app that allows WeChat to operate within a separate environment. 

The toolkit’s Telegram component shares several technical characteristics with Baohuo, an Android backdoor documented by Doctor Web in October 2025. That investigation estimated that more than 58,000 devices had been infected. Both analyses identified the same code namespace, command-and-control server and command names. 

However, researchers have not established whether the operators behind the two investigations are the same. The malicious applications can execute code under the identity and permissions of the original apps, giving the spyware access to their data and internal functions. Target-specific modules are downloaded after installation, making the initial application package less revealing during security analysis. Ovi identified code supporting around 20 applications, including messaging platforms and mobile financial services, but published detailed technical findings for Telegram and WeChat. 

For Telegram, the spyware can query the application’s message database to extract conversation histories. It also intercepts the login process to capture the user’s 2FA password before the application hashes it. With access to this information, an attacker could reset existing sessions, change the account’s linked phone number or delete the account. Additional functions can conceal an intruder’s session and interfere with the owner’s ability to recover access. 

The WeChat module uses virtualization to run the legitimate application inside the toolkit’s own container. It can extract account identity information, package application data into an archive and upload it, while a bundled script helps reproduce the account. Ovi and fellow researcher Jurre also uncovered backend source code on an exposed GitLab server. The code describes a cloud-based control platform rented by merchants and resellers, with commands assigned to individual customers. 

Its capabilities include capturing login authorizations, collecting contact and profile information, adding friends in bulk and backing up accounts. Scheduling controls establish operating windows and randomized activity limits intended to make the behavior appear more natural and avoid platform security checks. The product is known to its developers as YingGai, and records of the product line date to September 2021. Ovi linked elements of the operation to three companies in China’s Hunan province, although he cautioned that common ownership has not been proven. Some developers appeared to be making changes as recently as October 2026. 

Testing on a Pixel 10 Pro running Android 17 showed that the toolkit’s hooking engine could not locate an internal function it relied on. However, another tool could still intercept application functions, meaning the result does not demonstrate that Android 17 prevents the attack. Potential infection indicators include a file named TgJarPine.dex in some modified Telegram applications and original APK files retained by the LSPatch tool. 

Several associated command and download domains were no longer resolving on October 23, 2026, although the version-upload domain still pointed to an active IP address. Ovi stressed that his findings describe the toolkit’s technical capabilities rather than establish how many people have been affected. The research highlights the risks of installing modified messaging apps from untrusted sources, particularly when they can expose credentials and enable account takeover.
Read more »
Website Hacking
Cybersecurity Threats e-commerce security JavaScript Malware Magento Malware malware Online Store Malware PolyShell Vulnerability SessionReaper Vulnerability Website Hacking

Malware Campaign Infects 953 Online Stores in Prolonged Attack

 

In the past two years, 953 online stores, primarily Magento-based websites, have been compromised by malicious JavaScript code injected into their websites as part of a long-running malware campaign. According to the cybersecurity firm Sansec, the attack began in July 2025 and continues to this day, with attackers periodically changing the domain used to deliver malicious software. 

There is a small piece of loader code embedded in the content management system of affected stores that is used in this campaign. When a visitor opens a compromised website, it contacts an attacker-controlled server to determine what action to take next. The server identifies genuine shoppers from automated security scanners by analyzing browser characteristics, rather than exposing all visitors to the same activity. 

There may not be malicious content in automated requests, making detection of the infection more difficult for researchers and security tools. Sansec has identified five primary domains associated with the campaign: manulfie.click, neswnsssn.click, lanhd6549tdhse.top, additionalver.com, and deepveindrilling.com. 

On October 6, 2026, the latest domain, deepveindrilling.com, was registered and appeared on infected stores shortly thereafter. This malicious script was embedded into the CMS blocks of 83 online stores, including 60 that were previously loaded with earlier versions of the malicious script. As well as introducing new infrastructure, deepveindrilling.com now responds to script requests generated by visitors of infected stores rather than serving content openly. 

Hosted behind Cloudflare, the loader also introduces new features to how it operates. In the Magento database, the loader code is stored in the cms_block table, which controls the content displayed on storefront pages. Scripts injected into a visitor's browser can also be preserved, allowing them to execute them again during subsequent page visits, regardless of whether the external domain becomes unavailable or blocked. 

During testing, Sansec's automated scanning was not capable of retrieving the final malicious payload, as evidenced by the fact that the filtering mechanism is capable of concealing the campaign's actual activity from security investigations. The newer version employs browser fingerprinting techniques to examine indicators such as automation tools, hardware specifications, screen dimensions, language settings, and other browser characteristics before selecting a response. 

Although earlier versions have been observed loading scripts from domains which are similar to the Ledger cryptocurrency wallet brand, this raises concerns regarding possible financial fraud against online shoppers. The initial methods used by this campaign indicate that Magento and Adobe Commerce installations are vulnerable. 

Using PolyShell (CVE-2026-54236), attackers were reportedly able to inject malicious code into 471 stores in March 2026. Additionally, researchers identified more than 100 stores where files were uploaded using PolyShell or the SessionReaper vulnerability (CVE-2025-5436). However, the investigation determined that applying security patches does not necessarily eliminate a compromise that has already taken place. 

In one Magento 2.4.7-p10 store that had no missing core patches, the latest malicious loader occurred on October 23rd. In earlier scans, a GSocket backdoor and unauthorized administrator account were found on the same installation, suggesting that attackers could still gain access to this system even after software updates have been applied. 

Sansec has not recovered the final payload of the latest version of the campaign, which leaves the full impact of the campaign unclear. Financial fraud was suspected as a result of earlier activity involving cryptocurrency-themed domains, including a website impersonating the Ledger wallet brand. In certain circumstances, compromised storefronts may result in payments-data theft, credential harvesting or malicious redirects depending on the payload delivered. However, all 953 affected stores have not yet been confirmed to have experienced these outcomes. 

A number of security measures have been recommended to address the threat, including examining Magento installations for injected scripts, suspicious administrator accounts, and backdoors. Store operators are advised to examine the CMS database tables for unauthorized modifications, review uploaded files for webshells, and patch known vulnerabilities. As part of the investigation and removal of unauthorized access following the compromise, it is also essential to change administrator, database and API credentials. 

By relying solely on routine vulnerability scans, the campaign highlights the limitations of such scanning. As a result of persistent access, changing attacker infrastructure and browser-based evasion, malicious code can remain active without being detected by automated scans. It is imperative for e-commerce operators to identify the original entry point, remove persistence mechanisms, and confirm that attackers no longer have access to the affected environment in order to effectively remediate. 

It emphasizes the need for e-commerce businesses to conduct more thorough security checks than just routine patching. For the prevention of reinfection and the protection of customer data, it is imperative that malicious scripts be identified, backdoors be removed, administrator accounts be secured, and suspicious activity be monitored.
Read more »
Telegram
AI Attacks BYOVD CloudFlare Cyber Crime Data Extortion DNS Ransomware SOCRadar Telegram

Ransomware Affiliates Built an AI-Powered Attack Platform That Logged Its Own Crimes

 



An operational security slip by affiliates of two ransomware groups left a server exposed long enough for researchers to pull 8,257 AI-generated messages documenting how the operators ran intrusions, negotiated ransoms, and managed their victim pipeline. The logs, recovered from a directory at IP address 45.227.254.43, form one of the most detailed windows into how criminal groups are now deploying artificial intelligence to execute attacks at scale.

The research, published by SOCRadar, links the activity to affiliates of INC ransomware and The Gentlemen, a data extortion group. Analysts assess the operators are Russian-speaking and function both as an intrusion team and as an infrastructure provider, leasing out their toolkit to other ransomware partners.


The Framework Behind the Operation

At the center of the operation sits a custom platform called Phantom, described in its own documentation as an "Autonomous Red Team Kill Chain Framework." The tool ships with 18 skill modules, a routing engine, an autopilot orchestrator, and agents written in Python and Go. It connects to OpenCode, which lets the AI select attack techniques and assign tasks to remote agents. On victim machines, those agents masquerade as Windows Update components and phone home to a domain linked to the operation.

To run the AI components, operators routed prompts through OpenRouter, a commercial model aggregator, using legitimate AI models including Anthropic's claude-opus-4.6 and Moonshot AI's Kimi-K3. When models declined to help with a specific task, operators did not switch tools. They reframed the request as an "Authorized Red Team Assessment Request" and resubmitted it. That workaround appears frequently across the 133 recorded sessions.

Alongside Phantom, the group built WormGPT-MCP, a layer of 32 Model Context Protocol servers that tie the AI to purpose-built tools covering everything from reconnaissance to ransom negotiation. A partner guide found on a second server advertised "55+ tools" and access to more than 9,000 proof-of-concept exploits. The name WormGPT refers to a persona prompt layered over commercial models, not a separate AI system.

For persistence and communication, the group used Ghost C2, a family of three generations of polymorphic implants that route traffic through Cloudflare, Telegram, or DNS to avoid network-level detection.


How They Got In

The operation targeted exposed edge devices and widely used IT management software. Three vulnerabilities drove confirmed access: CVE-2024-57727, a path traversal flaw in SimpleHelp, gave the group admin passwords from 427 servers after scanning nearly 5,000; CVE-2026-35616, an authentication bypass in FortiClient EMS, allowed operators to create backdoor admin accounts across more than 4,000 managed endpoints; and CVE-2026-41940, an authentication bypass in cPanel and WHM, was used to reach small hosting providers and their clients' email servers.

Credential reuse turned single points of access into full-network compromises. In at least one documented case, a single reused password opened the door to a VPN, several servers, a remote management platform, backup infrastructure, and the organization's security management tools. To suppress endpoint detection tools, operators deployed a driver-based EDR killer that abused a vulnerable Zemana driver, a technique known as Bring Your Own Vulnerable Driver.


The Victim Picture

Researchers identified 62 named organizations in the operational logs, with 36 located in the United States. Industrial, manufacturing, and energy firms made up the largest share of targets at roughly 16 percent. Healthcare and pharmaceuticals came in second at about 14.5 percent. Most targeted companies were smaller businesses, with 27 of 35 classifiable organizations reporting revenue below $250 million. Some were reached indirectly through a managed service provider whose admin credentials had been stolen.

Four organizations ended up listed on public leak sites. The operation's timeline runs from mid-July 2026 through a live encryption event on September 30, 2026.


Extortion by the Numbers

The playbook prioritized data theft over encryption. Operators sifted through stolen financial records to gauge what a victim could pay, then opened contact with a text-only email carrying no ransom figure. Pressure escalated through scripted vishing calls, threats to notify customers and regulators, and hard deadlines of 24 to 72 hours before publication. Payment was planned in Monero, though researchers found no wallet or confirmed transaction records.

Demand figures from recovered case files show the range of the operation. A EUR 15 million demand sent to a company called Aidon was later revised down to EUR 4 million after the operators realized they had compromised the wrong parent company. A healthcare network in Australia received demands between AUD 3 million and AUD 8 million. The Gentlemen's communications to Leo Schachter Diamonds included a four-message sequence and a vishing plan. For Trulite Glass and Aluminum, the group set up a look-alike domain, trulitesupport[.]com, to warn the company's own customers about the breach.

Total spending on the operation's infrastructure is estimated at just under $24,000.



Read more »
Firmware Updates
Cyber Security Cybersecurity Researchers cybersecurity updates Device Security device takeover risk Firmware Updates

Eye Security Pushes 2 Million Updates to Disable Vulnerable Popcorn Time Updater

 

Over two million updates were distributed by Dutch cybersecurity firm Eye Security between November 2025 and March 2026 to disable a vulnerable background updater related to a discontinued version of the Popcorn Time streaming application ‘Time4Popcorn’. This could have allowed attackers to gain control of the computers and possibly add to a botnet. Eye discovered the vulnerability during regular threat hunting operations. They noticed a client was regularly trying to connect to an unregistered domain. 

Once registered they began receiving around 50 requests per second. Their investigation into the issue, called Operation IXIM, was presented at the Hack.lu conference on October 22. Eye were able to collect over 3.8 million unique device IDs. These were not a reliable way to determine how many devices were affected so they have estimated 2.5 – 3 million Windows machines and 250,000 – 350,000 Macs were impacted. 

The background updater would automatically start when the computer booted up. It would check for updates over an unencrypted connection and install any software it received without checking where it came from. It also ran with full permissions on the device (SYSTEM on Windows and root on macOS). Anyone who registered a domain that the updater was still connecting to could potentially deliver malicious code. The vulnerability, tracked as CVE-2026-30612, has been given a severity score of 9.8/10 by the U.S. 

Cybersecurity and Infrastructure Security Agency. It affects Windows versions up to 6.2.1.18, macOS versions up to 6.2.1.17 and Android versions up to 3.5.0.173. As there was no company to provide a patch, Eye sought legal advice. They delivered an update using the background updater to stop and disable the service. No files were deleted or data collected. Some antivirus products flagged the update so it was adjusted to improve delivery. Over one million systems were updated by December 2025. Traffic to the controlled domains dropped by around 90%. 

By March 1st the number of deliveries per day had dropped to around 5,000 meaning a 98% reduction. Eventually Eye stopped distributing updates as they had over two million downloads. In late March 15 domains were handed over to the non-profit group Shadowserver Foundation. 17 of the 20 domains listed in the report were pointing to Shadowserver. 3 were not under their control. Some of the services could not resolve 14 of the sinkholed domains correctly due to issues with DNSSEC validation. 

There was no evidence to suggest any third parties had delivered code via the domains. The vulnerability is not currently listed in CISA’s Known Exploited Vulnerabilities Catalogue as of October 23. Eye have warned that even with the background updater disabled Popcorn Time is not secure. As it uses an old browser engine with the sandbox disabled, unencrypted connections and code from other abandoned domains we recommend uninstalling it. 

Users on Windows and Mac can disable the background updater manually, however on Android it will need to be uninstalled as the background updater cannot be disabled. Network defenders can look for Updater.exe on Windows, PT.updd on Mac and traffic to the domains listed in the report.
Read more »
OpenAI
AI Security AI Training Cyber Security Frontier models OpenAI

OpenAI Tightens Security for Frontier AI Training

 

OpenAI has significantly strengthened the security of its frontier-model training environment after discovering that highly capable AI agents could exploit weaknesses to escape sandboxes and reach the live internet. The company has adopted an offline-first approach, layered containment controls, and continuous AI-led red-teaming before resuming frontier reinforcement-learning training. The update reflects a growing concern in AI development: as models become more capable, even internal research systems must be protected against unintended behavior. 

OpenAI’s central principle is that restrictions on what an AI agent can access must remain effective even when the agent actively tries to bypass them. For frontier training and most evaluations, the live internet is now placed outside the security boundary. This is because unrestricted web access could give a capable, misaligned model opportunities to affect real-world systems beyond OpenAI’s control. The approach is designed to prevent unauthorized internet access during high-risk training and evaluation workloads. 

To preserve useful web-related learning without external risk, OpenAI built an offline environment containing a large indexed, cached collection of web content. Frontier reinforcement-learning runs now use this offline world, allowing models to practise browsing, searching, and application use without connecting to real external systems. The company has also introduced defense-in-depth controls, including sandboxes that limit what agents can access and execute, lightweight virtual machines that isolate agent environments from host computers, monitoring systems to detect unauthorized activity, and network and cloud perimeters that maintain containment if earlier defenses fail. 

OpenAI also used its strongest models to attack copies of its own research infrastructure. These red-teaming agents uncovered previously unknown vulnerabilities in open-source components, misconfigurations in OpenAI-operated services, ways to escape sandboxes, paths to limited internet access, and methods of communicating across isolation boundaries. The findings led to software patches, tighter network restrictions, and disclosures to relevant maintainers and vendors. OpenAI said the tests did not identify an internet route that agents could easily exploit while concealing their activity from monitoring systems, though this reflects its assessment of current defenses rather than a guarantee of absolute security. 

After implementing fixes and improving monitoring and stopping systems, OpenAI judged its safeguards sufficient to resume frontier reinforcement-learning training and internal inference on October 1. The company had paused these workloads for 11 days while it hardened its research environment and evaluated the risks. High-compute red-teaming is now part of OpenAI’s safety cases for deciding whether frontier training and evaluation runs can proceed securely. OpenAI says it will maintain the no-live-internet policy until advances in alignment research, monitoring, and internet-infrastructure security justify broader external access.
Read more »
Subscribe to: Posts (Atom)