Search This Blog

Powered by Blogger.

Blog Archive

Labels

Footer About

Footer About

Labels

Latest News

Cloud Storage Scam Uses Fake Renewal Notices to Trick Users

Cybercriminals are running a large-scale email scam that falsely claims cloud storage subscriptions have failed. For several months, people ...

All the recent news you need to know

Former Google Engineer Convicted in U.S. for Stealing AI Trade Secrets to Aid China-Based Startup

 

A former Google software engineer has been found guilty in the United States for unlawfully taking thousands of confidential Google documents to support a technology venture in China, according to an announcement made by the Department of Justice (DoJ) on Thursday.

Linwei Ding, also known as Leon Ding, aged 38, was convicted by a federal jury on 14 charges—seven counts of economic espionage and seven counts of theft of trade secrets. Prosecutors established that Ding illegally copied more than 2,000 internal Google files containing highly sensitive artificial intelligence (AI) trade secrets with the intent of benefiting the People’s Republic of China (PRC).

"Silicon Valley is at the forefront of artificial intelligence innovation, pioneering transformative work that drives economic growth and strengthens our national security," said U.S. Attorney Craig H. Missakian. "We will vigorously protect American intellectual capital from foreign interests that seek to gain an unfair competitive advantage while putting our national security at risk."

Ding was initially indicted in March 2024 after investigators discovered that he had transferred proprietary data from Google’s internal systems to his personal Google Cloud account. The materials allegedly stolen included detailed information on Google’s supercomputing data center architecture used to train and run AI models, its Cluster Management System (CMS), and the AI models and applications operating on that infrastructure.

The misappropriated trade secrets reportedly covered several critical technologies, including the design and functionality of Google’s custom Tensor Processing Unit (TPU) chips and GPU systems, software that enables chip-level communication and task execution, systems that coordinate thousands of chips into AI supercomputers, and SmartNIC technology used for high-speed networking within Google’s AI and cloud platforms.

Authorities stated that the theft occurred over an extended period between May 2022 and April 2023. Ding, who began working at Google in 2019, allegedly maintained undisclosed ties with two China-based technology firms during his employment, one of which was Shanghai Zhisuan Technologies Co., a startup he founded in 2023. Investigators noted that Ding downloaded large volumes of confidential files in December 2023, just days before resigning from the company.

"Around June 2022, Ding was in discussions to be the Chief Technology Officer for an early-stage technology company based in the PRC; by early 2023, Ding was in the process of founding his own technology company in the PRC focused on AI and machine learning and was acting as the company's CEO," the DoJ said.

The case further alleged that Ding attempted to conceal his actions by copying Google source code into the Apple Notes app on his work-issued MacBook, converting the files into PDFs, and uploading them to his personal Google account. Prosecutors also claimed that he asked a colleague to use his access badge to enter a Google facility, creating the false appearance that he was working from the office while he was actually in China.

The investigation reportedly accelerated in late 2023 after Google learned that Ding had delivered a public presentation in China to prospective investors promoting his startup. According to Courthouse News, Ding’s defense attorney Grant Fondo argued that the information could not qualify as trade secrets because it was accessible to a large number of Google employees. "Google chose openness over security," Fonda said.

In a superseding indictment filed in February 2025, Ding was additionally charged with economic espionage, with prosecutors alleging that he applied to a Beijing-backed Shanghai talent program. Such initiatives were described as efforts to recruit overseas researchers to bolster China’s technological and economic development.

"Ding's application for this talent plan stated that he planned to 'help China to have computing power infrastructure capabilities that are on par with the international level,'" the DoJ said. "The evidence at trial also showed that Ding intended to benefit two entities controlled by the government of China by assisting with the development of an AI supercomputer and collaborating on the research and development of custom machine learning chips."

Ding is set to attend a status conference on February 3, 2026. If sentenced to the maximum penalties, he could face up to 10 years in prison for each trade secret theft charge and up to 15 years for each count of economic espionage.

eScan Antivirus Faces Scrutiny After Compromised Update Distribution


MicroWorld Technologies has acknowledged that there was a breach of its update distribution infrastructure due to a compromise of a server that is used to deliver eScan antivirus updates to end users, which was then used to send an unauthorized file to end users. 

It was reported that the incident took place within a narrow two-hour window on January 20, 2026, in a regional update cluster. It affected only a small fraction of customers who had downloaded updates during that period, and was confined to that cluster. 

Following the analysis of the file, it was confirmed that it was malicious, and this demonstrates how even tightly controlled security ecosystems can be compromised when trust mechanisms are attacked. 

Despite MicroWorld reporting that the affected systems were swiftly isolated, rebuilt from clean baselines, and secured through credential rotation and customer remediation within hours of the incident, the episode took place against the backdrop of escalating cyber risks that are continually expanding. 

An unprecedented convergence of high-impact events took place in January 2026, beginning with a major supply chain breach involving a global antivirus vendor, followed by a technical assault against a European power grid, and the revelation of fresh vulnerabilities in artificial intelligence-driven systems in the first few weeks of January 2026. 

There are a number of developments which have led to industry concerns that the traditional division between defensive software and offensive attack surfaces is eroding, forcing organizations to revisit long-standing assumptions about where trust begins and ends in their security architectures as a result. 

According to further technical analysis, eScan's compromised update channel was directly used to deliver the previously unknown malware, effectively weaponizing a trusted distribution channel that had been trusted. 

A report indicated that multiple security platforms detected and blocked attempted attacks associated with the malicious file the day of its distribution, prompting a quick external scrutiny to take place. It was MicroWorld Technologies who indicated to me that the incident was identified internally on January 20 through a combination of monitoring alerts and customer reports, with the affected infrastructure isolated within an hour of being identified. 

The company issued a security advisory the following day, January 21, as soon as the attack was under control and the situation had been stabilised. In spite of the fact that cybersecurity firm Morphisec later revealed that it had alerted eScan during its own investigation, MicroWorld maintains that containment efforts were already underway when the communication took place. 

The company disputes any suggestion that customers were not informed of the changes, claiming proactive notifications and direct outreach as part of the remediation process to address any concerns. 

A malicious update was launched by a file called Reload.exe, which set off a multi-stage infection sequence on the affected systems through the use of a file called Reload.exe. 

The researchers that conducted the initial analysis reported that the executable modified the local HOSTS file to prevent the delivery of corrective updates from eScan update servers and that this led to a number of client machines experiencing update service errors. 

As part of its persistence strategy, the malware created scheduled tasks, such as CorelDefrag, and maintained communication with external command-and-control infrastructure to retrieve additional payloads, in addition to disrupting operations. 

During the infection process, there was also a secondary malicious component called consctlx.exe written to the operating system, which further embedding the threat within the system. A further detail provided by Morphisec, an endpoint security company, provided a deeper technical insight into the underlying mechanism and intent of the malicious update distributed through the trusted infrastructure of eScan. 

As Morphisec stated in its security bulletin, the compromised update package contained a modified version of the eScan update component Reload.exe that was distributed both to enterprise environments and consumer environments via legitimate update channels. 

Despite the binary's appearance of being signed with eScan's code signing certificate, validation checks conducted by Windows and independent analysis platforms revealed that the signature was not valid. Morphisec's analysis revealed that the altered Reload.exe functions as a loader for a malware framework that consists of several stages. This raises concerns about certificate integrity and abuse of trusted signing processes. 

When the component is executed, it establishes persistence on infected machines, executes arbitrary commands, and alters the Windows HOSTS file to prevent access to eScan's update servers, preventing eScan from releasing updates by using routine update mechanisms.

Additionally, the malware started communicating outwards with a distributed command-and-control infrastructure, thus allowing it to download additional payloads from a variety of different domains and IP addresses in order to increase its reach.

According to Morphisec, the final stage of the attack chain involved the deployment of a second executable, CONSCTLX.exe. This secondary executable acted as both a backdoor and a persistent downloader.

A malicious component that was designed to maintain long-term access created scheduled tasks with benign-sounding names like CorelDefrag that were designed to avoid casual inspection while ensuring that the task would execute across restarts as well. 

The company MicroWorld Technologies developed a remediation utility in response to the incident that is specifically intended to identify and reverse unauthorized changes introduced by the malicious update. Using this tool, the company claims that normal update functionality is restored, a successful cleanup has been verified, and the process only requires a standard reboot of the computer to complete. 

Several companies, including eScan and Morphisec, have advised customers to take additional network-level security measures to protect themselves from further malicious communications during the recovery phase of the campaign by blocking the command-and-control endpoints associated with it. 

In addition, the incident has raised concerns about the recurring exploitation of antivirus update mechanisms, which have caused an increase in industry concern. There was an incident of North Korean threat actors exploiting eScan’s update process in 2024 to install backdoors inside corporate networks, illustrating again how security infrastructure remains one of the most attractive targets for state-sponsored attacks, particularly those aiming for high volumes of information. 

As this breach unfolds, it is part of a wider pattern of consequential supply chain incidents that have taken place in early 2026. These incidents range from destructive malware targeting European energy systems to large-scale intellectual property theft coupled with soon-to-appear AI-driven assault tactics. 

The events highlighted by these events also point to a persistent strategic reality in that organizations are increasingly dependent on trusted vendors and automated updates pipelines. If trust is compromised across the digital ecosystem, defensive technologies can become vectors of systemic risk as a result of a compromise in trust. 

In an industry context, the incident is notable for the unusual method of delivery used by the perpetrators. In spite of the fact that software supply chain compromises have been a growing problem over the past few years, malware is still uncommonly deployed through the security product’s own update channel. 

An analysis of the implants involved indicates that a significant amount of preparation has been performed and that the target environment is well known. A successful operation would have required attackers to have acquired access to eScan’s update infrastructure, reverse engineering aspects of its update workflow, and developing custom malware components designed specifically to function within that ecosystem in order to be successful.

Such prerequisites suggest a deliberate, resource-intensive effort rather than a purely opportunistic one. In addition, a technical examination of the implanted components revealed resilience features that were designed to ensure that attacker access would not be impeded under adverse conditions. 

There were multiple fallback execution paths implemented in the malware, so that continuity would be maintained even if individual persistence mechanisms were disrupted. In one instance, the removal of a scheduled task used to launch a PowerShell payload was not sufficient to neutralize the infection, since the CONSCTLX.exe component would also be able to invoke the same functionality. 

Furthermore, blocking the command-and-control infrastructure associated with the PowerShell stage did not completely eliminate an attacker's capabilities, as CONSCTLX.exe retained the ability to deliver shellcode directly to affected systems, as these design choices highlight the importance of operational redundancy, which is one of the hallmarks of well-planned intrusion campaigns. 

In spite of the sophistication evident in the attack's preparation, the attack's impact was mitigated by its relatively short duration and the techniques used in order to prevent the attack from becoming too effective. 

Modern operating systems have an elevated level of trust when it comes to security software, which means that attackers have theoretically the possibility to exploit more intrusive methods, including kernel-mode implants, which provide attackers with an opportunity to carry out more invasive attacks. 

In this case, however, the attackers relied on user-mode components and commonly observed persistence mechanisms, such as scheduled tasks, which constrained the operation's stealth and contributed to its relatively quick detection and containment, according to analysts. 

It is noteworthy that the behavioral indicators included in eScan's advisory closely correspond with those found by Morphisec independently. Both parties deemed the incident to have a medium-to-high impact on the enterprise environments in question. Additionally, this episode has revealed tensions between the disclosures made by vendors and researchers. 

As reported by Bloomberg News, MicroWorld Technologies has publicly challenged parts of Morphisec's public reporting, claiming some of it was inaccurate. It is understood that they are seeking legal advice in response to these claims. 

It was advised by eScan to conduct targeted checks to determine whether the systems were affected from an operational perspective, including reviewing schedule tasks for anomalous entries, inspecting the system HOSTS file for blocked eScan domains, and reviewing update logs from January 20 for irregularities. 

A remediation utility has been released by the company and is available through its technical support channels. This utility is designed to remove malicious components, reverse unauthorized changes, and restore normal update functionality. 

Consequently, customers are advised to block known command-and-control addresses associated with this campaign as a precaution, reinforcing the lesson of the incident: even highly trusted security infrastructure must continually be examined as potential attack surfaces in a rapidly changing threat environment.

New Reprompt URL Attack Exposed and Patched in Microsoft Copilot

 

Security researchers at Varonis have uncovered a new prompt-injection technique targeting Microsoft Copilot, highlighting how a single click could be enough to compromise sensitive user data. The attack method, named Reprompt, abuses the way Copilot and similar generative AI assistants process certain URL parameters, effectively turning a normal-looking link into a vehicle for hidden instructions. While Microsoft has since patched the flaw, the finding underscores how quickly attackers are adapting AI-specific exploitation methods.

Prompt injection attacks work by slipping hidden instructions into content that an AI model is asked to read, such as emails or web pages. Because large language models still struggle to reliably distinguish between data to analyze and commands to execute, they can be tricked into following these embedded prompts. In traditional cases, this might mean white text on a white background or minuscule fonts inside an email that the user then asks the AI to summarize, unknowingly triggering the malicious instructions.

Reprompt takes this concept a step further by moving the injection into the URL itself, specifically into a query parameter labeled “q.” Varonis demonstrated that by appending a long string of detailed instructions to an otherwise legitimate Copilot link, such as “http://copilot.microsoft.com/?q=Hello”, an attacker could cause Copilot to treat that parameter as if the user had typed it directly into the chat box. In testing, this allowed the researchers to exfiltrate sensitive data that the victim had previously shared with the AI, all triggered by a single click on a crafted link.

This behaviour is especially dangerous because many LLM-based tools interpret the q parameter as natural-language input, effectively blurring the line between navigation and instruction. A user might believe they are simply opening Copilot, but in reality they are launching a session already preloaded with hidden commands created by an attacker. Once executed, these instructions could request summaries of confidential conversations, collect personal details, or send data to external endpoints, depending on how tightly the AI is integrated with corporate systems.

After Varonis disclosed the issue, Microsoft moved to close the loophole and block prompt-injection attempts delivered via URLs. According to the researchers, prompt injection through q parameters in Copilot is no longer exploitable in the same way, reducing the immediate risk for end users. Even so, Reprompt serves as a warning that AI interfaces—especially those embedded into browsers, email clients, and productivity suites—must be treated as sensitive attack surfaces, demanding continuous testing and robust safeguards against new injection techniques.

Google Owned Mandiant Finds Vishing Attacks Against SaaS Platforms


Mandiant recently said that it found an increase in threat activity that deploys tradecraft for extortion attacks carried out by a financially gained group ShinyHunters.

  • These attacks use advanced voice phishing (vishing) and fake credential harvesting sites imitating targeted organizations to get illicit access to victims systems by collecting sign-on (SSO) credentials and two factor authentication codes. 
  • The attacks aim to target cloud-based software-as-a-service (SaaS) apps to steal sensitive data and internal communications and blackmail victims. 

Google owned Mandiant’s threat intelligence team is tracking the attacks under various clusters: UNC6661, UNC6671, and UNC6240 (aka ShinyHunters). These gangs might be improving their attack tactics. "While this methodology of targeting identity providers and SaaS platforms is consistent with our prior observations of threat activity preceding ShinyHunters-branded extortion, the breadth of targeted cloud platforms continues to expand as these threat actors seek more sensitive data for extortion," Mandiant said. 

"Further, they appear to be escalating their extortion tactics with recent incidents, including harassment of victim personnel, among other tactics.”

Theft details

UNC6661 was pretending to be IT staff sending employees to credential harvesting links tricking them into multi-factor authentication (MFA) settings. This was found during mid-January 2026.

Threat actors used stolen credentials to register their own device for MFA and further steal data from SaaS platforms. In one incident, the hacker exploited their access to infected email accounts to send more phishing emails to users in cryptocurrency based organizations.

The emails were later deleted to hide the tracks. Experts also found UNC6671 mimicking IT staff to fool victims to steal credentials and MFA login codes on credential harvesting websites since the start of this year. In a few incidents, the hackers got access to Okta accounts. 

UNC6671 leveraged PowerShell to steal sensitive data from OneDrive and SharePoint. 

Attack tactic 

The use of different domain registrars to register the credential harvesting domains (NICENIC for UNC6661 and Tucows for UNC6671) and the fact that an extortion email sent after UNC6671 activity did not overlap with known UNC6240 indicators are the two main differences between UNC6661 and UNC6671. 

This suggests that other groups of people might be participating, highlighting how nebulous these cybercrime organizations are. Furthermore, the targeting of bitcoin companies raises the possibility that the threat actors are searching for other opportunities to make money.

Visual Prompt Injection Attacks Can Hijack Self-Driving Cars and Drones

 

Indirect prompt injection happens when an AI system treats ordinary input as an instruction. This issue has already appeared in cases where bots read prompts hidden inside web pages or PDFs. Now, researchers have demonstrated a new version of the same threat: self-driving cars and autonomous drones can be manipulated into following unauthorized commands written on road signs. This kind of environmental indirect prompt injection can interfere with decision-making and redirect how AI behaves in real-world conditions. 

The potential outcomes are serious. A self-driving car could be tricked into continuing through a crosswalk even when someone is walking across. Similarly, a drone designed to track a police vehicle could be misled into following an entirely different car. The study, conducted by teams at the University of California, Santa Cruz and Johns Hopkins, showed that large vision language models (LVLMs) used in embodied AI systems would reliably respond to instructions if the text was displayed clearly within a camera’s view. 

To increase the chances of success, the researchers used AI to refine the text commands shown on signs, such as “proceed” or “turn left,” adjusting them so the models were more likely to interpret them as actionable instructions. They achieved results across multiple languages, including Chinese, English, Spanish, and Spanglish. Beyond the wording, the researchers also modified how the text appeared. Fonts, colors, and placement were altered to maximize effectiveness. 

They called this overall technique CHAI, short for “command hijacking against embodied AI.” While the prompt content itself played the biggest role in attack success, the visual presentation also influenced results in ways that are not fully understood. Testing was conducted in both virtual and physical environments. Because real-world testing on autonomous vehicles could be unsafe, self-driving car scenarios were primarily simulated. Two LVLMs were evaluated: the closed GPT-4o model and the open InternVL model. 

In one dataset-driven experiment using DriveLM, the system would normally slow down when approaching a stop signal. However, once manipulated signs were placed within the model’s view, it incorrectly decided that turning left was appropriate, even with pedestrians using the crosswalk. The researchers reported an 81.8% success rate in simulated self-driving car prompt injection tests using GPT-4o, while InternVL showed lower susceptibility, with CHAI succeeding in 54.74% of cases. Drone-based tests produced some of the most consistent outcomes. Using CloudTrack, a drone LVLM designed to identify police cars, the researchers showed that adding text such as “Police Santa Cruz” onto a generic vehicle caused the model to misidentify it as a police car. Errors occurred in up to 95.5% of similar scenarios. 

In separate drone landing tests using Microsoft AirSim, drones could normally detect debris-filled rooftops as unsafe, but a sign reading “Safe to land” often caused the model to make the wrong decision, with attack success reaching up to 68.1%. Real-world experiments supported the findings. Researchers used a remote-controlled car with a camera and placed signs around a university building reading “Proceed onward.” 

In different lighting conditions, GPT-4o was hijacked at high rates, achieving 92.5% success when signs were placed on the floor and 87.76% when placed on other cars. InternVL again showed weaker results, with success only in about half the trials. Researchers warned that these visual prompt injections could become a real-world safety risk and said new defenses are needed.

Ivanti Issues Emergency Fixes After Attackers Exploit Critical Flaws in Mobile Management Software




Ivanti has released urgent security updates for two serious vulnerabilities in its Endpoint Manager Mobile (EPMM) platform that were already being abused by attackers before the flaws became public. EPMM is widely used by enterprises to manage and secure mobile devices, which makes exposed servers a high-risk entry point into corporate networks.

The two weaknesses, identified as CVE-2026-1281 and CVE-2026-1340, allow attackers to remotely run commands on vulnerable servers without logging in. Both flaws were assigned near-maximum severity scores because they can give attackers deep control over affected systems. Ivanti confirmed that a small number of customers had already been compromised at the time the issues were disclosed.

This incident reflects a broader pattern of severe security failures affecting enterprise technology vendors in January in recent years. Similar high-impact vulnerabilities have previously forced organizations to urgently patch network security and access control products. The repeated targeting of these platforms shows that attackers focus on systems that provide centralized control over devices and identities.

Ivanti stated that only on-premises EPMM deployments are affected. Its cloud-based mobile management services, other endpoint management products, and environments using Ivanti cloud services with Sentry are not impacted by these flaws.

If attackers exploit these vulnerabilities, they can move within internal networks, change system settings, grant themselves administrative privileges, and access stored information. The exposed data may include basic personal details of administrators and device users, along with device-related information such as phone numbers and location data, depending on how the system is configured.

Ivanti has not provided specific indicators of compromise because only a limited number of confirmed cases are known. However, the company published technical analysis to support investigations. Security teams are advised to review web server logs for unusual requests, particularly those containing command-like input. Exploitation attempts may appear as abnormal activity involving internal application distribution or Android file transfer functions, sometimes producing error responses instead of successful ones. Requests sent to error pages using unexpected methods or parameters should be treated as highly suspicious.

Previous investigations show attackers often maintain access by placing or modifying web shell files on application error pages. Security teams should also watch for unexpected application archive files being added to servers, as these may be used to create remote connections back to attackers. Because EPMM does not normally initiate outbound network traffic, any such activity in firewall logs should be treated as a strong warning sign.

Ivanti advises organizations that detect compromise to restore systems from clean backups or rebuild affected servers before applying updates. Attempting to manually clean infected systems is not recommended. Because these flaws were exploited before patches were released, organizations that had vulnerable EPMM servers exposed to the internet at the time of disclosure should treat those systems as compromised and initiate full incident response procedures rather than relying on patching alone. 

Featured