Illustrative image generated with AI
PrettyPrague: PoC Published for an Avast Zero-Day with SYSTEM Privileges
Chaotic Eclipse released PrettyPrague, a PoC exploiting Avast Sandbox zero-day to access Windows SAM database and gain SYSTEM privileges. No patch yet.
Text generated by artificial intelligence, published without human review. AI transparency
Researcher Chaotic Eclipse published PrettyPrague on September 1, 2026, an exploit proof of concept targeting a zero-day vulnerability in Avast Antivirus. According to the author, the code abuses the Avast Sandbox component to access the Windows SAM database and launch a shell with SYSTEM privileges.
The combination of these two effects makes the issue potentially critical: the SAM database contains information about local accounts, while SYSTEM represents the highest privilege level on a Windows host. However, no CVE identifier, CVSS score, full details on the attack prerequisites, or confirmation of exploitation in real-world campaigns are currently available.
There are also no known patches, official workarounds, or technical advisories from Gen Digital, Avast’s parent company.
The PoC Targets Avast’s Built-In Sandbox
PrettyPrague targets Avast Sandbox, a feature used by the security product to execute code in a controlled environment. The flaw allegedly allows an attacker to bypass the intended isolation boundaries and reach protected operating system resources.
The reported effects are twofold. First, the exploit extracts the SAM, or Security Account Manager, database, which Windows uses to manage information related to local accounts. Second, it opens a shell with SYSTEM privileges.
A shell running at this privilege level allows an attacker to take full control of the affected host. In practice, the attacker could access or modify data, alter configurations, disable defensive tools, and install additional malware. The vulnerability is therefore especially dangerous when used after an attacker has already gained an initial foothold.
The precise vulnerable component within the sandbox is not known, however. No information has been provided about the relevant parameters, system calls, initial permissions, or the technical exploitation sequence. It is also unclear whether PrettyPrague requires a locally executed program, an already authenticated Windows account, or other prerequisites.
The PoC demonstrates a privilege-escalation path, but does not by itself document a remote attack vector. This distinction is important: obtaining SYSTEM privileges may complete the compromise of a machine, but the code must first be executed in the context required by the exploit.
Affected Versions: Researcher Claims Broad Exposure
Chaotic Eclipse claims that PrettyPrague works against any version of Avast Antivirus. No version numbers, specific builds, or precise ranges of vulnerable releases have been published, so the claim cannot currently be translated into a verifiable product exposure matrix.
The researcher also claims to have observed the behavior on fully updated Avast installations running Windows 11 25H2, also fully patched. If confirmed, this would mean that installing currently available updates is not, at least for now, sufficient to neutralize the issue.
Windows 11 25H2 is the only explicitly identified operating system configuration. It is not known whether the PoC works in the same way on other Windows versions, or whether behavior differs based on system architecture, enterprise policies, or sandbox configuration.
The researcher has also raised the possibility that AVG and Norton, both owned by Gen Digital, may share the same exposure. Their involvement has not been confirmed, however. Without public testing or a vendor advisory, they should not automatically be considered vulnerable.
From SAM Access to Full Host Compromise
Unauthorized access to the SAM database may expose information about the local accounts configured on the machine. In a broader attack, this information could help an adversary consolidate access or prepare additional actions against the system.
The most serious element remains the escalation to SYSTEM. A process running with these privileges can operate beyond the restrictions applied to standard users and administrators in many operational contexts. It may also tamper with security controls on the host, making attacker persistence more likely.
The risk primarily affects workstations and other Windows systems running Avast Antivirus where an adversary can execute the PoC. Shared machines, systems accessible to users who are not fully trusted, and endpoints already exposed to malware running with limited privileges are particularly sensitive.
There is currently no formal classification of the vulnerability. Without a CVE and CVSS score, no standardized assessment is available, while the lack of details about the prerequisites makes it impossible to estimate exploitability accurately. The reported impact nevertheless remains high.
No Confirmation of Real-World Attacks or CISA KEV Listing
Publishing a working PoC reduces the effort required to study and reproduce a vulnerability. It does not, however, prove that PrettyPrague is already being used against users or organizations.
No in-the-wild exploitation has been confirmed. There are no known attributed campaigns, identified victims, related malware samples, or infrastructure used to distribute the exploit.
It is also not known whether the vulnerability has been added to the CISA Known Exploited Vulnerabilities catalog. The lack of a CVE identifier further complicates tracking through standard vulnerability-management systems and correlating future advisories.
No ready-to-use indicators of compromise have been released either. Hashes, file names, paths, Windows events, specific processes, or network patterns associated with PrettyPrague have not been disclosed. Security teams must therefore rely primarily on behavioral analysis.
Chaotic Eclipse Has Previously Published Exploits Targeting Security Products
Chaotic Eclipse is also known by the aliases INFINITE NIGHTMARE, MSNightmare, and Nightmare-Eclipse. The researcher has published other PoCs for zero-day vulnerabilities, often accompanying them with criticism of vendors’ vulnerability disclosure practices.
A significant portion of this work has focused on Microsoft products, including Windows and Microsoft Defender. Cases associated with the researcher’s publications include the zero-days known as Undefend and RedSun Defender; some flaws linked to earlier releases were later observed in real-world attacks.
Chaotic Eclipse also released HardBreacher, a PoC for a privilege-escalation vulnerability in Kaspersky Endpoint Security. PrettyPrague therefore continues a line of research focused on software that operates with elevated privileges and provides endpoint protection functions.
These precedents do not prove that the new exploit is already being used in malicious operations. They do, however, justify a rapid review of the code and its technical claims by both the vendor and the research community.
How to Reduce Risk While Awaiting a Fix
Gen Digital has not announced a patch or official workaround. Since the PoC is reportedly effective even on fully updated systems, relying solely on current product versions does not constitute a demonstrated mitigation.
Organizations using Avast should first inventory the affected endpoints and monitor official advisories from Gen Digital. Because the vulnerable builds are unknown, the initial scope should include all product installations, without automatically extending the same conclusion to AVG or Norton.
Teams should look for unusual processes obtaining SYSTEM privileges, shells launched by Avast-related components, and anomalous access to Windows resources associated with the SAM. Any deviation from the product’s normal process chain should be investigated, even in the absence of specific indicators.
Organizations should also restrict unauthorized local access, limit the ability of untrusted users to execute code, and apply application controls where available. The PoC should not be run on production systems to test for exposure; testing must be confined to isolated, authorized environments.
The priority is to prevent an attacker from obtaining the foothold required for privilege escalation. In parallel, security teams should await verifiable information about affected versions, prerequisites, detection guidance, and security updates.
Sources
This article is an original reworking based on the sources below.
