Amazon Web Services warns users of a vulnerability affecting TorchServe – a tool used by some of the largest companies in the world in the creation of artificial intelligence models in their business.
The technology giant published an opinion On the bug on Monday, CVE-2023-43654, and urged customers to update the latest version of Torcherve in order to solve the problem, which essentially exposes important administrative tools to the Internet.
The CVE-2023-43654 is part of a set of vulnerabilities called “Shelltorch” by researchers from the Israeli security company Oligo, which discovered problems.
Cob is a popular Open source code package in the Pytorch Ecosystem, which is supervised by Amazon and Meta. The project is used by hundreds of organizations from around the world, including companies like Walmart, Openai, Tesla, Azure, Google Cloud and Intel.
Using the vulnerabilities discovered by Oligo, a hacker could visualize, modify, steal or delete AI models and sensitive data that move between the company and the TorchServe server, according to the researchers.
Oligo has published details on another bug – CVE -2022-1471 – as well as a problem related to API configuration errors.
Researchers Idan Levcovich, Guy Kaplan and Gal Elbaz said that using an IP scanner, they had discovered “thousands of vulnerable cases exposed publicly, including some of the largest organizations in the world – open to Unauthorized access and insertion of malicious AI models, and potentially full access and insertion of malicious AI models, and potentially a complete and potentially full access and insertion of malicious AI models, and potential Take -Over server. “”
They noted the popularity of the Pytorch in research on automatic learning as well as the AI projects of private companies.
“This is why he shocked our researchers to discover that – without any authentication – we could carry out the code remotely with high privileges, using new critical vulnerabilities in servers of models with Open Pytorch (TorchServe)”, said researchers Oligo. “These vulnerabilities make it possible to compromise servers around the world. As a result, some of the world’s largest companies could be at immediate risk. »»
Neither AWS nor Oligo said the vulnerabilities were exploited. Oligo Created a free tool That organizations can use to see if they are affected by the problem. The researchers said Meta and Amazon have published updates that tackle some of the problems.
Researchers have also provided other advice to businesses, including the reconfiguration of management consoles and limiting access to trusted areas. Meta did not respond to requests for comments.
The problem occurs a few days after two other popular open source libraries – libvpx And libwebp – have turned out to have vulnerabilities exploited by pirates.
Oligo noted that TorchServe vulnerabilities highlight the serious dangers associated with artificial intelligence models that are based strongly on open source software.
The White House and a handful of government agencies have called for experts to help them create policies around the cybersecurity of open source software and to promote the use of more secure programming languages.
They held a summit last month on the issue and published a roadmap To find out how the deep causes of open source problems can be treated in the future.
Callie Guenther, principal director of cyber-menace research at the Cybersecurity Company, Critical Start, told Recred Future News that it is now essential that IA models being widely used in the academic world and industry are not armed as vectors for exploits.
CVE-2023-43654, the most serious of vulnerabilities according to experts, “accentuates the need for white list mechanisms rigorously tested. An “authorized list” which accepts without discrimination all areas is, paradoxically, a flagrant security gap, “she said.
The other vulnerability – CVE -2022-1471 – is a well -known problem, explained Guenther. The fact that it is present in a tool as popular as TorchServe was another example of “the importance of in -depth security journals, in particular when the existing libraries is implemented”.
“Since TorchServe has the support of industry giants like Meta and Amazon and is widely used in the technological sector, such vulnerabilities can collapse in a myriad of applications, compromising the integrity of AI models And affiliated systems, ”she said.
She added that technology giants should be more proactive about the use of third-party security assessments to catch problems like these earlier.
Clarification (10/4/2023): This story has been updated to reflect that the opinion was published by AWS, an Amazon unit.