Nvidia patches critical Triton server bugs that threaten AI model security

Threat Score:
70
CSO Online
6 hours ago

Overview

A surprising attack chain in Nvidia’s Triton Inference Server, starting with a seemingly minor memory-name leak, could allow full remote server takeover without user authentication. Security researchers from Wiz have discovered a chain of critical vulnerabilities in the popular open-source platform for running AI models at scale. “When chained together, these flaws can potentially allow a remote, unauthenticated attacker to gain complete control of the server, achieving remote code execution (RC...

Related Articles

5 articles
4

Microsoft Launches Zero-Day Quest Hacking Contest with Rewards Up to $5 Million

GB Hackers • 2 hours ago

Microsoft Launches Zero-Day Quest Hacking Contest with Rewards Up to $5 Million Microsoft has unveiled the return of its groundbreaking Zero Day Quest initiative, escalating the stakes in cybersecurity research with a staggering total bounty pool of up to $5 million. Building on the success of last year’s inaugural event, which offered $4 million in awards and garnered overwhelming participation from the global security community, this year’s program intensifies focus on high-impact vulnerabilit

Score
84
Read more
5

APT36 Targets Indian Government: Credential Theft Campaign Uncovered

GB Hackers • 5 hours ago

APT36 Targets Indian Government: Credential Theft Campaign Uncovered A sophisticated phishing campaign attributed with medium confidence to the Pakistan-linked APT36 group, also known as Transparent Tribe or Mythic Leopard, has been uncovered targeting Indian defense organizations and government entities. This operation employs typo-squatted domains that mimic official Indian government platforms, such as mail.mgovcloud.in and virtualeoffice.cloud, to deceive users into surrendering credentials.

Score
82
Read more