Hijacking Vertex AI Model Uploads for Cross-Tenant RCE


Palo Alto Unit42 discovered a vulnerability in the Google Cloud Vertex AI software development kit (SDK) for Python, and responsibly disclosed it to Google. Before Google’s fix, the vulnerability would have allowed an attacker operating entirely from their own Google Cloud project to hijack a victim’s model upload and poison it. By exploiting this flaw in vulnerable versions of the SDK, an attacker can achieve remote code execution (RCE) within a target’s Vertex AI serving infrastructure, with zero initial access to the victim’s project.

Read more…
Source:  Palo Alto Unit 42


Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox


Related:

  • Unpatched Apple Zero-Day in macOS Finder Allows Code Execution

    September 22, 2021

    A zero-day security vulnerability in Apple’s macOS Finder system could allow remote attackers to trick users into running arbitrary commands, according to researchers – and a silent patch hasn’t fixed it. For those not in the Apple camp, the macOS Finder is the default file manager and GUI front-end used on all Macintosh operating systems. It’s ...

  • Microsoft MSHTML Flaw Exploited by Ryuk Ransomware Gang

    September 17, 2021

    Criminals behind the Ryuk ransomware were early exploiters of the Windows MSHTML flaw, actively leveraging the bug in campaigns ahead of a patch released by Microsoft this week. Collaborative research by Microsoft and RiskIQ revealed campaigns by Ryuk threat actors early on that exploited the flaw, tracked as CVE-2021-40444. The bug is a remote code execution ...

  • An American Company Fears Its Windows Hacks Helped India Spy On China And Pakistan

    September 17, 2021

    Earlier this year, researchers at Russian cybersecurity firm Kaspersky witnessed a cyberespionage campaign targeting Microsoft Windows PCs at government and telecom entities in China and Pakistan. They began in June 2020 and continued through to April 2021. What piqued the researchers’ interest was the hacking software used by the digital spies, whom Kaspersky had dubbed ...

  • Exploitation of the CVE-2021-40444 vulnerability in MSHTML

    September 16, 2021

    Last week, Microsoft reported the remote code execution vulnerability CVE-2021-40444 in the MSHTML browser engine. According to the company, this vulnerability has already been used in targeted attacks against Microsoft Office users. In attempt to exploit this vulnerability, attackers create a document with a specially-crafted object. If a user opens the document, MS Office will ...

  • Azure Zero-Day Flaws Highlight Lurking Supply-Chain Risk

    September 16, 2021

    Four Microsoft zero-day vulnerabilities in the Azure cloud platform’s Open Management Infrastructure (OMI) — a software that many don’t know is embedded in a host of services — show that OMI represents a significant security blind spot, researchers said. Collectively dubbed “OMIGOD” because of the name and the reaction of the researchers who discovered them, the ...

  • APT Actors Exploiting Newly Identified Vulnerability in ManageEngine ADSelfService Plus

    September 16, 2021

    This joint advisory is the result of analytic efforts between the Federal Bureau of Investigation (FBI), United States Coast Guard Cyber Command (CGCYBER), and the Cybersecurity and Infrastructure Security Agency (CISA) to highlight the cyber threat associated with active exploitation of a newly identified vulnerability (CVE-2021-40539) in ManageEngine ADSelfService Plus—a self-service password management and single ...