Chaotic Eclipse discloses MiniPlasma zero-day, suggesting a missing or undone 2020 Windows security fix


Chaotic Eclipse discloses MiniPlasma zero-day, suggesting a missing or undone 2020 Windows security fix

Pierluigi Paganini
May 18, 2026

MiniPlasma: a Windows SYSTEM privilege escalation believed patched in 2020 (CVE-2020-17103) is still fully working on every patched Windows 11.

Once again, security researcher Chaotic Eclipse has released a proof-of-concept exploit for a new Windows privilege escalation zero-day called MiniPlasma, which can grant attackers SYSTEM privileges on fully patched systems.

The flaw affects “cldflt.sys,” the Windows Cloud Files Mini Filter Driver, specifically within the “HsmOsBlockPlaceholderAccess” routine. Google Project Zero researcher James Forshaw originally reported the vulnerability to Microsoft in September 2020.

“After re-investigating the technique used in GreenPlasma (specifically SetPolicyVal), it turns out cldflt!HsmOsBlockPlaceholderAccess is still vulnerable to the exact same issue that was reported to Microsoft 6 years ago. I’m not taking full credit for this, James Forshaw from google project zero found the vulnerability and reported it to Microsoft and was supposedly fixed as CVE-2020-17103.” Chaotic Eclipse wrote.

“However, a research who’s a friend of mine pointed out that the routine might still have a vulnerability, which is something I considered but brushed off because I thought it was impossible for Microsoft to just not patch this or rollback the patch.”

Chaotic Eclipse investigated further and found that the exact same vulnerability is still present in fully patched systems running the latest May 2026 updates. The original proof-of-concept code published by Forshaw worked without modification. The researcher then weaponized it to spawn a SYSTEM shell and published it as MiniPlasma, noting that reliability may vary due to the exploit’s race-condition nature, but that it worked consistently across their test environments.

“After investigating, it turns out the exact same issue that was reported to Microsoft by Google project zero is actually still present, unpatched. I’m unsure if Microsoft just never patched the issue or the patch was silently rolled back at some point for unknown reasons. The original PoC by Google worked without any changes.” Chaotic Eclipse added. “To highlight this issue, I weaponized the original PoC to spawn a SYSTEM shell. It seems to work reliably in my machines but success rate may vary since it’s a race condition. I believe all Windows versions are affected by this vulnerability.”

Will Dormann, a popular cybersecurity researcher, independently confirmed the result: MiniPlasma opens a cmd.exe prompt with SYSTEM privileges on Windows 11 running the latest patches. He noted it does not work on the Insider Preview Canary build, which suggests Microsoft may be addressing it there, but that provides little comfort to the hundreds of millions of users running production Windows 11 builds.

“New from Nightmare-Eclipse, we have MiniPlasma [github.com]Dormann wrote. “Works reliably to get a SYSTEM cmd.exe prompt on Win11 (including 26H1) with May’s updates. Is reportedly a failure to properly fix CVE-2020-17103 [msrc.microsoft.com]. I’ll note that it does not seem to work on the latest Insider Preview Canary Windows 11.”

Mysteriously, a patch reportedly confirmed in 2020 appears to have disappeared. The issue goes beyond delayed updates and raises broader concerns about the reliability and completeness of Windows patch management, leaving organizations questioning whether fully patched systems are truly secure.

But now, let’s focus on Chaotic Eclipse.

There is a GitHub profile called Nightmare-Eclipse. Behind it, a researcher who goes by Chaotic Eclipse. In the span of a few weeks, this individual has published working exploit code for five separate Windows vulnerabilities, some previously unknown, some believed to have been patched years ago but apparently still very much alive. The disclosures triggered a wave of zero-days that put Microsoft under pressure, raised concerns about the reliability of its patches, and revived the long-running debate over whether publishing exploit code promotes transparency or creates greater security risks.

To understand the significance of what Chaotic Eclipse has published, it helps to lay out the full picture of what has been disclosed so far.

The first two flaws in the Defender series, BlueHammer, RedSun, and UnDefend, appeared in April. BlueHammer and RedSun let attackers escalate privileges locally in Microsoft Defender. UnDefend instead triggers a denial-of-service, blocking security definition updates and weakening protection. Microsoft addressed BlueHammer as CVE-2026-33825, but RedSun and UnDefend remained unpatched. Within days of the public release, Huntress researchers observed real-world exploitation of all three. Attackers began using BlueHammer on April 10, then moved to the proof-of-concept code for RedSun and UnDefend on April 16, following the publicly available exploit code with a precision that left little doubt about where the attack playbook had come from.

Then came YellowKey and GreenPlasma, two more Windows zero-days disclosed by the same researcher and reported by Security Affairs. YellowKey is a BitLocker bypass issue that affects Windows 11 and Windows Server 2022/2025 systems. The flaw allows attackers with physical access to bypass BitLocker protections and gain unrestricted shell access to encrypted volumes through the Windows Recovery Environment (WinRE). The attack is triggered by placing specially crafted files inside a specific directory on a USB drive or directly in the EFI partition. What makes this flaw particularly unsettling is not just its functionality but also the researcher’s commentary on its origins: the vulnerable component exists exclusively within the WinRE image, not in standard Windows installations, and an identical component appears in normal installations but without the triggering functionality.

Chaotic Eclipse drew an uncomfortable conclusion from this: “Now why would I say this is a backdoor ? The component that is responsible for this bug is not present anywhere (even in the internet) except inside WinRE image and what makes it raise suspicions is the fact that the exact same component is also present with the exact same name in a normal windows installation but without the functionalities that trigger the bitlocker bypass issue. Why ? I just can’t come up with an explanation beside the fact that this was intentional. Also for whatever reason, only windows 11 (+Server 2022/2025) are affect, windows 10 is not.

It is a claim that Microsoft has not publicly addressed. Whether it reflects a genuine design anomaly, an architectural oversight that looks suspicious from the outside, or something else entirely, is not known. What is known is that the flaw works, that it affects Windows 11 and Server 2022/2025, and that Windows 10 is not affected, a distinction that itself raises questions without obvious answers.

The second flaw in that pair, GreenPlasma, targets the Windows Collaborative Translation Framework — the CTFMON subsystem, and enables privilege escalation on Windows 11 and Windows Server 2022/2026 by creating arbitrary memory section objects inside directories writable by SYSTEM. The researcher withheld the full exploit chain but noted that someone with the right skills could complete the escalation from the published material. A partial disclosure that is effectively complete for a skilled attacker is a category that sits uncomfortably between responsible and irresponsible release.

Who Is Chaotic Eclipse?

Tracing a precise profile of the researcher is difficult. They operate under a pseudonym, maintain a GitHub repository under the handle Nightmare-Eclipse, and communicate through a blog and occasional social media posts. The documentation accompanying each release, while not exhaustive, reflects genuine understanding of the underlying mechanisms.

The motivations behind Chaotic Eclipse are not fully clear, but public comments point to frustration with Microsoft’s patching process, a concern shared by many in the security community. By publishing working exploit code instead of following standard coordinated disclosure timelines, the researcher seems to be pushing for faster action and greater accountability. This view is reinforced by the fact that some flaws were quickly patched after public exposure, while others remained unaddressed.

There is also a more serious concern raised in the disclosures, including the possibility that one issue may reflect intentional design rather than a simple vulnerability. Whether or not that is accurate, it shows a more confrontational approach to disclosure than traditional reporting channels.

Overall, this reflects a long-standing divide in security research: some researchers work within vendor programs and disclosure frameworks, while others publish findings directly. The latter approach can pressure companies into fixing issues faster, but it also risks exposing users to active attacks before patches are ready.

The MiniPlasma situation really shows how divided the vulnerability disclosure debate still is, because both sides actually have a point.

Chaotic Eclipse’s argument is based on something concrete: Microsoft originally fixed CVE-2020-17103 back in 2020, yet parts of that fix now seem to be missing in newer Windows builds. Without a public disclosure and a working proof of concept, it’s fair to ask whether the issue would have been noticed at all. We’ve already seen something similar with BlueHammer. Once exploit details became public, Microsoft reacted quickly and pushed out fixes, which suggests that public attention can sometimes force action faster than private reporting alone.

At the same time, the risks of releasing exploit code are very real. Researchers at Huntress have repeatedly pointed out that attackers move fast once proof-of-concept code is available. In some cases, weaponization happens within days. That creates a difficult tradeoff: public research helps defenders and increases pressure on vendors, but it also gives threat actors a shortcut. Even if MiniPlasma is not trivial to exploit consistently, the fact that Windows runs on billions of devices means that any reliable exploit immediately becomes high risk.

That tension is exactly why responsible disclosure became standard practice in cybersecurity. Typically, researchers report vulnerabilities privately, vendors get time to investigate and release a fix, and technical details are published afterward. The process is not perfect, and sometimes vendors move too slowly, but the goal is to reduce unnecessary exposure while still ensuring accountability. Chaotic Eclipse’s approach speeds that timeline up considerably. That can lead to faster patching, but it also reduces the time defenders have before attackers begin experimenting with the same information.

What makes the MiniPlasma story more concerning, though, is not just the disclosure debate itself. The bigger issue is the possibility that a vulnerability fixed years ago may have quietly reappeared. If a patch released in 2020 can effectively disappear because of regressions, refactoring, or build changes, then it challenges a basic assumption many organizations rely on: that once something is patched, it stays fixed.

That matters because modern enterprise security depends heavily on patch management. Microsoft ships hundreds of fixes every year, and security teams generally trust that updates permanently close known holes. If old vulnerabilities can unintentionally return over time, then patching alone is no longer enough. Teams may also need ways to continuously verify that protections are still present after later updates and feature changes.

The fact that related issues have reportedly appeared multiple times in the same component only adds to that concern. From a defender’s perspective, MiniPlasma is less about one exploit and more about what it says regarding software maintenance at scale. It highlights the gap between how patching is supposed to work in theory and how difficult it can be to guarantee in practice across years of development and constant code changes.

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, MiniPlasma)







Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


Embodied Intelligence and the Phenomenology of AI explores how human cognition arises from perception, embodiment, and experience in contrast to disembodied artificial intelligence.

Conceptual diagram illustrating embodied intelligence and the phenomenology of AI through perception, embodiment, environment, and experience.

A Conscious Intelligence Perspective

The rapid development of artificial intelligence has transformed modern discussions about cognition and intelligence. Machine learning systems now recognize patterns in data, generate language, analyze images, and assist with complex decision-making processes across scientific, economic, and technological domains. These capabilities have led some observers to suggest that artificial systems may eventually replicate or even surpass human intelligence.

Yet beneath these technological achievements lies a fundamental philosophical question: what does it mean to be intelligent? While artificial intelligence can perform impressive computational tasks, human cognition emerges from a far more complex interaction between perception, embodiment, and lived experience. Understanding this distinction requires examining the concept of embodied intelligence—the idea that human cognition arises through the dynamic interaction between mind, body, and environment.

Phenomenology, the philosophical study of conscious experience, offers a powerful framework for understanding embodied intelligence. Rather than treating cognition as a purely abstract computational process, phenomenology emphasizes that perception, thought, and understanding occur within a lived world shaped by sensory experience and bodily engagement. When applied to contemporary discussions of artificial intelligence, this perspective reveals important differences between human cognition and machine intelligence.

Within the framework of Conscious Intelligence (CI), embodied intelligence highlights the experiential foundations of human awareness and interpretation. It underscores why human cognition remains essential in guiding technological systems, particularly as artificial intelligence continues to expand its capabilities.

Understanding Embodied Intelligence

The concept of embodied intelligence challenges traditional views of cognition that treat the mind as an abstract information-processing system. Early models of artificial intelligence often assumed that intelligence could be replicated through symbolic reasoning and computational logic. According to this perspective, cognition could be understood as the manipulation of symbols according to formal rules.

However, research in cognitive science and philosophy has increasingly shown that human intelligence cannot be separated from bodily experience. Perception, movement, and environmental interaction play fundamental roles in shaping how individuals understand the world (Varela, Thompson, & Rosch, 1991).

Embodied intelligence suggests that cognition arises through continuous engagement between the organism and its environment. Rather than operating as a detached reasoning system, the mind develops within the context of sensory perception and physical action.

Consider a simple example: observing a bird in flight. This experience involves more than visual pattern recognition. The observer’s body subtly adjusts posture, attention tracks motion through space, and prior experiences shape expectations about movement and behavior. The act of perception becomes an integrated process involving vision, spatial awareness, memory, and anticipation.

This dynamic interaction between perception and action forms the basis of embodied cognition. Intelligence emerges not from isolated computation but from the ongoing relationship between body and world.

Phenomenology and the Lived Body

Phenomenology provides a philosophical foundation for understanding embodied intelligence. While early phenomenologists such as Edmund Husserl explored the intentional structure of consciousness, later thinkers emphasized the central role of the body in shaping perception and cognition.

The French philosopher Maurice Merleau-Ponty argued that human consciousness is fundamentally embodied. In his influential work Phenomenology of Perception, he described the body as the primary site through which individuals encounter the world (Merleau-Ponty, 2012). Rather than functioning as an object separate from consciousness, the body becomes the medium through which experience unfolds.

According to Merleau-Ponty, perception is not merely the passive reception of sensory data. Instead, it is an active process in which the body engages with the environment through movement, orientation, and attention. The body provides a framework through which space, time, and meaning become intelligible.

This perspective challenges purely computational models of intelligence. Artificial systems may process visual data or recognize objects in images, but they do not experience the world through a lived body. They do not move within environments, feel spatial relationships, or engage with objects through physical interaction.

Phenomenology therefore highlights a crucial distinction between human cognition and artificial intelligence: human intelligence is grounded in embodied experience, while most AI systems operate within abstract computational environments.

The Limits of Disembodied Artificial Intelligence

Modern artificial intelligence systems excel at tasks involving pattern recognition and data analysis. Deep learning networks can identify faces in images, translate languages, and predict complex trends based on large datasets. These capabilities have created the impression that machine intelligence may soon approximate human cognition.

However, AI systems typically operate in disembodied informational spaces. They process data within computational architectures rather than through physical interaction with the world. Their “perception” consists of numerical representations rather than lived sensory experience.

Philosopher Hubert Dreyfus argued that early AI research underestimated the importance of embodied and contextual knowledge in human cognition (Dreyfus, 1992). Humans navigate the world through intuitive understanding shaped by years of bodily interaction with their environment. Much of this knowledge remains implicit rather than formally articulated.

For example, people can effortlessly grasp objects, maintain balance while walking, or recognize subtle emotional expressions in social interactions. These abilities arise from complex sensorimotor systems that integrate perception and action.

Replicating such capabilities in artificial systems has proven extraordinarily challenging. While robotics research has made significant progress, the embodied adaptability of biological organisms remains difficult to reproduce through purely computational methods.

This limitation suggests that human intelligence involves dimensions of cognition that extend beyond algorithmic processing. Embodied experience provides a context for understanding that cannot easily be reduced to data structures or symbolic reasoning.

Embodiment and Meaning

One of the most important implications of embodied intelligence concerns the nature of meaning. Human understanding emerges through interaction with environments that are experienced through the body.

Language, for example, is deeply connected to embodied experience. Words describing spatial relationships, movement, and sensation reflect how humans encounter the world physically. Even abstract concepts often originate from metaphors grounded in bodily perception.

Artificial intelligence systems can generate language that appears coherent and meaningful, yet they do not experience the embodied contexts that give language its significance. Large language models predict patterns in textual data without possessing an experiential understanding of the concepts they describe.

This distinction helps explain why AI systems sometimes produce outputs that appear plausible yet lack deeper comprehension. Without embodied experience, machines cannot anchor meaning in lived reality.

Phenomenology therefore emphasizes that understanding involves more than symbolic manipulation. Meaning arises from engagement with the world, shaped by perception, movement, and social interaction.

Embodied Intelligence in Human Practice

Embodied intelligence is visible in many aspects of human activity. Artists, athletes, musicians, and craftspeople rely heavily on forms of knowledge that cannot easily be articulated through formal rules. Their expertise develops through repeated interaction between perception and action.

In observational practices such as photography, for example, perception involves more than simply recording visual information. The observer anticipates movement, adjusts bodily orientation, and interprets environmental cues to capture meaningful moments. These processes occur through embodied awareness rather than through explicit calculation.

Scientific inquiry also involves embodied intelligence. Researchers conduct experiments, manipulate instruments, and interpret physical phenomena through sensory engagement with experimental environments. Knowledge emerges through interaction between theory, observation, and experience.

These examples illustrate how intelligence unfolds through embodied practice. Human cognition develops not only through abstract reasoning but also through lived engagement with the world.

Embodied Intelligence and Conscious Intelligence

Within the framework of Conscious Intelligence, embodiment plays a crucial role in shaping how individuals understand and guide technological systems. The CI model emphasizes three pillars—meta-awareness, interpretive agency, and responsible alignment—and embodied intelligence provides experiential grounding for each.

Meta-awareness involves reflecting on one’s own cognitive processes. Phenomenological reflection encourages individuals to examine how perception and bodily engagement influence understanding.

Interpretive agency arises from the human capacity to assign meaning to experiences. Embodied perception provides the contextual richness that allows individuals to interpret information within lived environments.

Responsible alignment involves directing technological capabilities toward ethical and constructive purposes. Embodied awareness can deepen ethical reflection by highlighting the real-world consequences of technological decisions for human experience.

By emphasizing embodiment, the CI framework reinforces the importance of human awareness in guiding artificial intelligence. Machines may extend computational capabilities, but human cognition provides the experiential perspective necessary to interpret and apply technological outputs responsibly.

Toward Embodied Artificial Intelligence

Recognizing the limitations of disembodied AI has led some researchers to explore the possibility of embodied artificial intelligence. Robotics and sensorimotor learning systems attempt to integrate perception and action within physical environments.

These approaches acknowledge that intelligence may require interaction with the world rather than purely abstract computation. Robots equipped with sensors and mobility can learn through environmental feedback, gradually developing adaptive behaviors.

While such research represents an important step toward more flexible AI systems, replicating the complexity of human embodiment remains a significant challenge. Biological organisms possess highly sophisticated sensory systems, neural architectures, and evolutionary adaptations that enable nuanced interactions with their surroundings.

Nevertheless, the exploration of embodied AI highlights an important philosophical insight: intelligence may be inseparable from the environments in which it develops.

Embodied Intelligence in a Technological Civilization

As artificial intelligence becomes increasingly integrated into modern societies, understanding embodied intelligence becomes more important than ever. Digital technologies shape how individuals perceive information, communicate with others, and interact with the world.

Yet human cognition continues to depend on embodied experience. Perception, movement, and sensory engagement remain essential components of understanding.

The rise of AI therefore does not eliminate the importance of human intelligence. Instead, it emphasizes the need for conscious awareness capable of interpreting technological systems within lived contexts.

Embodied intelligence reminds us that cognition is not simply an abstract computational function. It is an activity embedded in perception, experience, and interaction with the world.

Conclusion

The concept of embodied intelligence reveals a fundamental dimension of human cognition often overlooked in discussions of artificial intelligence. While machines excel at processing data and recognizing patterns, human intelligence arises through the dynamic interaction between mind, body, and environment.

Phenomenology provides a philosophical framework for understanding this relationship by examining the structures of lived experience. Through the work of thinkers such as Merleau-Ponty, phenomenology shows that perception and understanding emerge from embodied engagement with the world.

In the age of artificial intelligence, this perspective becomes increasingly relevant. AI systems may extend human analytical capabilities, but they remain fundamentally different from human cognition, which is grounded in embodied experience.

Within the framework of Conscious Intelligence, embodied intelligence underscores the importance of human awareness in guiding technological systems. By integrating reflection, interpretation, and responsibility, individuals can ensure that artificial intelligence serves constructive purposes within human societies.

Ultimately, understanding intelligence requires acknowledging the role of the body in shaping perception and meaning. Human awareness remains rooted in lived experience, and this experiential foundation continues to guide the evolving relationship between human cognition and artificial intelligence.

References

Dreyfus, H. L. (1992). What computers still can’t do: A critique of artificial reason. MIT Press.

Merleau-Ponty, M. (2012). Phenomenology of perception. Routledge. (Original work published 1945)

Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. MIT Press.



Source link