Over the weekend, hackers were able to break into Nvidia's systems and steal roughly 1TB worth of confidential files. So far, the group behind the attack claims to have source code, drivers, an LHR bypass, schematics and more and now, some of these files have begun leaking – starting with the source code for DLSS.
The group, known as LAPSUS$, has demanded that Nvidia drops performance limitations from GeForce graphics cards and releases an update to bypass LHR on RTX 30 series GPUs. So far, Nvidia has not been in contact with the hacking group, so the group is apparently beginning to sell off the stolen data.
In the image above, which was passed along to TechPowerUp via an anonymous tip, we can see what appears to be the source code for DLSS. Deep Learning Super Sampling is a proprietary technology from Nvidia, which leverages the Tensor Cores on RTX GPUs to reconstruct low resolution images and present them in higher detail, thus saving performance in games and leading to significantly higher frame rates.
The version of DLSS included in the leak is v2.2, which is a fairly recent version of the feature. If the files end up being widely spread, it opens the door for Nvidia's rivals to have a poke around, as well as for the open source community to potentially bring DLSS to other platforms like Linux. However, it is also worth noting that some of the information around how DLSS works is already public knowledge through the DLSS SDK, which is freely available on GitHub.
Discuss on our Facebook page, HERE.
KitGuru Says: Nvidia has confirmed the cyberattack at this point and is continuing to investigate. Whether or not the potential for major leaks pushes the company to negotiate with the attackers still remains to be seen.