Trend Micro Open Sources Cybertron LLM for Cybersecurity
Trend Micro today announced it will open source a Cybertron large language model (LLM) specifically trained to automate a wide range of cybersecurity tasks.
Additionally, Trend Micro plans to make available a series of artificial intelligence (AI) agents under an open-source license, starting with one to perform cloud risk analysis tasks.
Rachel Jin, chief enterprise product officer for Trend Micro, said the Trend Cybertron model, including weights, will be made available under a permissive MIT license to both encourage adoption and contributions from across the cybersecurity community.
Unveiled last month, Cybertron is based on version 3.1 of the open-source Llama LLM originally developed by Meta and is currently optimized to be deployed using the NVIDIA Inference Microservices (NIM) framework that makes it simpler to deploy artificial intelligence (AI) models.
Trend Micro trained Cybertron using its documentation and cybersecurity expertise without exposing customer data, to ensure the LLM does not inadvertently share any proprietary information, noted Jin. It is currently based on an 8-billion-parameter AI model, with a 70-billion-parameter version planned to provide more advanced reasoning capabilities to enable AI agents to automate more complex tasks.
The overall goal is to reduce the total cost of developing AI capabilities for cybersecurity use cases, with vendors such as Trend Micro focusing instead on how to include these and other AI capabilities in an integrated Trend Micro One cloud service that automates a wider range of cybersecurity tasks and workflows, added Jin.
It’s not clear how many cybersecurity vendors might be inclined to standardize on a potentially narrow set of open-source LLMs they customize versus investing in training their own. However, in theory, relying more on open-source AI models should reduce total costs in a way that could be passed on to customers.
Fernando Montenegro, vice president and practice lead for cybersecurity for The Futurum Group, said as cybersecurity continues to evolve the AI capabilities being provided will only continue to become more robust. Open source AI models for cybersecurity, as such, are a welcome development, he noted.
Ultimately, AI will reduce overall levels of fatigue by reducing false positives, in addition to making it possible to focus remediation efforts on the most lethal threats. That’s critical because given the limited amount of cybersecurity expertise available many organizations currently limit the number of investigations they conduct.
Exactly how AI will transform cybersecurity remains to be seen, but a lot of the toil that inevitably burns cybersecurity professionals out should be sharply reduced. It’s not likely AI will eliminate the need for cybersecurity professionals any time soon but it should help even a playing field that today is decidedly lopsided.
Of course, cybercriminals will similarly leverage AI to launch more sophisticated attacks faster than ever, so in that sense, the cybersecurity industry is now caught up in an AI arms race. However, by relying more on open-source AI models the cost of the arms race might not be nearly as high as many might have initially anticipated.
