New Vulnerability Targets Open Source AI Giant Ray Leaving Thousands Of Companies Exposed

Attackers are said to be targeting popular AI framework giant Ray by exposing it to a vulnerability that has no cure yet.

This puts thousands of companies at risk of having their data exposed in regards to AI workload as per a report published by Oligo Security recently.

The flaw forces attackers to grab hold of firms involved in computing power as it could leak bundles of sensitive data. And it’s not something new, as explained by experts. It actually dates back several months and impacts sectors such as crypto, biopharma, and even education.

Such ordeals were first seen in the latter part of 2023 combined with a host of other serious flaws. And some are still getting patched while a few like this one seem to have no cure as a whole so far.

So many teams are still not aware of how exposed their systems are to this vulnerability because it’s not showing up on their respective static scans. Whatever the case may be, it’s still causing a serious breach and major loss.

So many servers of Ray were exposed and continue to be compromised as well. In such circumstances, we see hackers abusing the system to download crypto software.

Ray is a giant platform that leading tech names like Amazon, Uber, and even OpenAI continue to use. They’re great in terms of speed as well as efficiency. Models including the likes of GPT-4 are compromising the system as they make use of huge computational power.

These types of models cannot be made themselves fit across a single machine’s memory. But this Ray technology is the way most of the models are run. Within very little time, it transformed into one of the best practices in the industry, especially in regards to those making the most of AI like those experts of Python as well as other GPUs and a host of series of machines.

The servers get compromised after leaking a lot of data that’s dubbed sensitive. And as per reports from several leading researchers, the workload from AI producers continues to be compromised.

While experts of the study in question failed to specify which companies were impacted as a result of this, they did unveil how many companies arising from various sectors were getting affected.
There was also plenty of evidence on display in regards to the rollout of DB credentials as well as private SSH keys and even some tokens related to OpenAI. This would allow hackers to easily access OpenAI accounts without the need for credentials.

All GPU models that continue to be compromised daily aren’t in stock right now and they’re quite hard to grab a hold of for obvious reasons, not to mention very costly too.

The total costs linked to machines and power computation are also getting compromised on the way and estimated costs could go up one one billion dollars, as predicted by experts.

So the first hint of the attack taking place was September of last year and that gave such culprits a whopping amount of time to compromise the hardware and achieve their gains without anyone noticing.

There’s a lot more to the matter as explained in the technical blog published by Oligo so we do advise you to take a look in case you want to know more about this attack and what sorts of mitigation strategies are in place.

Image: DIW-Aigen

Read next: Meta’s Oversight Board Says The Company’s Ban On The Arabic Term ‘Martyr’ Restricts Free Expression
Previous Post Next Post