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:

  • Here’s how we got persistent shell access on a Boeing 747 – Pen Test Partners

    May 21, 2021

    Researchers from infosec biz Pen Test Partners established a persistent shell on an in-flight entertainment (IFE) system from a Boeing 747 airliner after exploiting a vulnerability dating back to 1999. It’s an attack that’s more of a curiosity than anything else: it’s too difficult to pull off during an actual flight, and it’s rare these days ...

  • Open Source Vulnerabilities Converging DevOps & SecOps

    May 20, 2021

    Workplace evolution is in favor of traditional siloes being torn down and replaced with increased cross-functional collaboration, working in lockstep to deliver better outcomes. But it is not as easy as it sounds. Security and development teams have historically worked in siloes, which has created a long- standing disconnect between them. Both teams are responsible for ...

  • Hackers scan for vulnerable devices minutes after bug disclosure

    May 19, 2021

    Every hour, a threat actor starts a new scan on the public web for vulnerable systems, moving at a quicker pace than global enterprises when trying to identify serious vulnerabilities on their networks. The adversaries’ efforts increase significantly when critical vulnerabilities emerge, with new internet-wide scans happening within minutes from the disclosure. Read more… Source: Bleeping Computer  

  • Windows PoC Exploit Released for Wormable RCE

    May 19, 2021

    A researcher has released a proof-of-concept (PoC) exploit for CVE-2021-31166, a use-after-free, highly critical vulnerability in the HTTP protocol stack (http.sys) that could lead to wormable remote code execution (RCE). Microsoft discovered the flaw internally, releasing a patch in its May 11 Patch Tuesday update. This was the most severe bug in that batch: an http.sys ...

  • May Android security updates patch 4 zero-days exploited in the wild

    May 19, 2021

    According to info provided by Google’s Project Zero team, four Android security vulnerabilities were exploited in the wild as zero-day bugs before being patched earlier this month. Attacks attempting to exploit these flaws were targeted and impacted a limited number of users based on information shared after this month’s Android security updates were published. Read more… Source: Bleeping ...

  • Stalkerware Apps Riddled with Security Bugs

    May 18, 2021

    Android stalkerware apps – used to surreptitiously track people’s movements and digital activities – turn out to themselves be rife with security holes that put victims in even danger. Stalkerware can track the GPS location of a victim’s device, record conversations, capture images and snoop on browser histories. And overall, according to ESET researcher Lukas Stefanko, ...