Attackers Chain JFrog Artifactory Flaws to Seize Admin Control

Attackers have exploited a series of critical vulnerabilities in self-hosted JFrog Artifactory servers to gain full administrator control. These breaches compromise the heart of software build pipelines, creating a high-risk vector for supply chain attacks.

According to reports, attackers utilized two distinct methods to seize control. In one scenario, a chain of two vulnerabilities—CVE-2026-42018 and CVE-2026-42016—was used to first leak an anonymous-user token and then swap it for a token with administrator scope. Security firm Wiz observed these chained attacks occurring between August 15 and September 8, 2026. Simultaneously, a separate critical authentication bypass, CVE-2026-82329, provided a more direct path to admin access. This flaw carries a CVSS score of 9.8, and Fastly recorded approximately 406,000 exploitation attempts of this specific vulnerability on September 2, 2026, alone.

The Supply Chain Risk

JFrog Artifactory serves as a central repository where software builds pull necessary dependencies. Because it sits at the core of the development lifecycle, administrator access is catastrophic. Once inside, attackers installed malicious Groovy plugins to achieve remote code execution (RCE), created unauthorized administrator accounts, and deployed Rust-based backdoors to maintain persistence. Some of these vulnerabilities were originally identified during security evaluations conducted by researchers at OpenAI.

Industry Implications

The primary danger of these exploits is the potential for downstream contamination. By controlling the Artifactory server, an attacker can inject malicious code into the dependencies used by an organization, effectively poisoning every piece of software produced by that company. While JFrog stated that its cloud-hosted instances were not affected, the risk to self-hosted environments remains severe due to the nature of the persistence mechanisms used.

Remediation Challenges

Security experts warn that simply applying a software patch may not be sufficient to secure a compromised system. Fastly noted that a patch does not revoke tokens already minted, meaning attackers who have already stolen administrator tokens may retain access even after the vulnerability is closed. Organizations must not only patch their servers but also audit for unauthorized admin accounts and rotate compromised cluster join keys to fully evict intruders. Security teams are advised to monitor for the presence of unauthorized Groovy plugins and unusual Rust binaries on their servers.