Summary
- Wiz has confirmed real-world exploitation of three Artifactory vulnerabilities across multiple environments.
- Observed activity includes attacker-created administrator accounts, malicious Groovy plugins, command execution, web shells, and Rust backdoors.
- CISA now lists all three vulnerabilities as known exploited, materially strengthening the evidence since Cyber Insider's earlier Artifactory report.
Attackers exploiting vulnerabilities in JFrog Artifactory have progressed from initial access to administrative takeover and persistent compromise, according to new telemetry that substantially expands the evidence around attacks first reported earlier this month.
Wiz Research says it has confirmed in-the-wild exploitation of CVE-2026-42016, CVE-2026-42018, and CVE-2026-82329 across multiple self-hosted environments. Attackers have chained the first two vulnerabilities to bypass access restrictions and obtain administrative privileges, while other actors have successfully exploited the critical CVE-2026-82329 authentication bypass.
The US Cybersecurity and Infrastructure Security Agency has also placed all three vulnerabilities in its Known Exploited Vulnerabilities catalogue. CVE-2026-82329 was added on 2 September, while CVE-2026-42016 and CVE-2026-42018 were added on 11 September.
That represents a significant development from the initial Artifactory exploitation reports, when the strongest public evidence around CVE-2026-82329 came from external authorities and security reporting rather than detailed post-compromise telemetry.
Wiz has now observed attacker-created administrator accounts used for persistence, malicious Groovy plugins deployed for code execution, operating-system commands used for reconnaissance and file enumeration, second-stage payload delivery, web shells, and custom Rust backdoors communicating with command-and-control infrastructure.
The researchers stress that multiple actors are involved and that no single attacker performed every observed action. The findings describe a range of post-exploitation behaviour seen across compromised systems rather than one uniform campaign.
Artifactory’s position in software delivery gives those actions wider significance. The platform is used to store and distribute packages, binaries, container images, build artefacts, and other components that organisations rely on during development and release processes.
Administrative access can therefore expose more than the repository host. Depending on configuration, an attacker may reach proprietary software, credentials, integrations, deployment information, or artefacts that downstream systems and developers treat as trusted.
Persistence makes incident handling more complicated again. Applying the vendor patch may close the original vulnerability while leaving behind an attacker-created administrator, malicious plugin, web shell, backdoor, or altered repository content. A successfully exploited server has to be treated differently from one that was merely vulnerable.
The three flaws create different routes into the platform. JFrog describes CVE-2026-82329 as an authentication weakness that can give an unauthenticated network attacker administrative privileges under default configurations. Wiz says CVE-2026-42018 can expose an internal anonymous-user token, while CVE-2026-42016 can be used to escalate the scope of a low-privileged token.
Wiz’s telemetry also shows uneven patching. It reported that remediation of the critical CVE-2026-82329 accelerated after disclosure, but that a substantial proportion of observed organisations remained vulnerable two weeks later. The lower-severity flaws showed slower remediation despite their ability to be chained.
That is a recurring weakness in vulnerability management. Individual CVSS scores can obscure what becomes possible when flaws are combined, particularly in infrastructure that holds trusted software and credentials. Two weaknesses that look less consequential in isolation may produce administrative access when placed in the same attack path.
Software repositories sit between development and production, making their integrity part of the wider software supply chain. A compromised repository can affect not only the system itself but the confidence placed in artefacts moving through it.
The evidence around Artifactory has therefore changed materially. The question is no longer whether exploitation reports are credible; CISA and Wiz both establish real-world use. The remaining task is determining which exposed systems were actually compromised and whether persistence or downstream artefact manipulation occurred before they were remediated.
