The Revolution of AI Security by Cybertron’s Evolution into “Cyber Brain”

By seamlessly integrating with Trend Vision One, Trend Cybertron introduces a unified and intelligent security solution that not just defends against existing threats but also proactively predicts and prevents upcoming security obstacles. This proactive strategy, along with its broad coverage spanning email, network, endpoint, and AI security fields, positions it as a crucial element in the security architecture of every contemporary enterprise.

This cohesive intelligence system exemplifies Trend Micro’s dedication to harnessing state-of-the-art AI technology to offer extensive, forward-looking security solutions that shield organizations against present and emerging threats in the increasingly intricate digital realm.

For users of Trend Vision One, Trend Cybertron seamlessly integrates with the platform, empowering users to effortlessly access the benefits within an integrated cybersecurity platform.

In What Way Does Trend Cybertron Differ within the Open-Source Environment?

At Trend Micro, we place our belief in the strength of community. That’s why we are thrilled to share specific models, datasets, and agents of Trend Cybertron with the open-source society. By promoting cooperation and inventive thinking in AI-oriented cybersecurity, we invite you to become part of a larger endeavor.

The various components of Trend Cybertron Open-Source are accessible via three primary channels:

1. NVIDIA NIM Catalog (https://build.nvidia.com/):
   • Tailored model utilizing Llama 3.1 and fine-tuned for deployment with NVIDIA NIM inference microservices
   • Incorporates an 8-billion-parameter AI model with plans for an advanced 70-billion parameter version
   • Showcases multiple AI agent blueprints for smooth integration and automation of security tasks
   • Empowers organizations to enhance security stance, reduce alert overload, save developer resources, and extract higher value from existing security sensors
2. Hugging Face (https://huggingface.co/trend-cybertron): Highlighted datasets and models from Trend Cybertron are part of the Primus collection and incorporate:

   Models:

   • trend-cybertron/Llama-Primus-Base: A text generation model derived from Llama-3.1-8B-Instruct, continuously pre-trained on 2.77B cybersecurity text tokens, achieving a 15.88% boost across cybersecurity benchmarks
   • trend-cybertron/Llama-Primus-Merged: A text generation model focused on instruction-following, maintaining capabilities of Llama-3.1-8B-Instruct while achieving a 14.84% advancement across cybersecurity benchmarks
   • trend-cybertron/Llama-Primus-Reasoning: A text generation model distilled on reasoning and reflection data for cybersecurity tasks, exhibiting a 10% improvement on CISSP (Certified Information Systems Security Professional) practice exams compared to humans

   Datasets:

   • trend-cybertron/Primus-Seed: High-quality cybersecurity texts meticulously gathered from reliable sources like Wikipedia, MITRE, cybersecurity company websites, and CTI
   • trend-cybertron/Primus-FineWeb: 2.57B tokens of cybersecurity texts sifted from FineWeb
   • trend-cybertron/Primus-Instruct: Roughly 1K QA pairs covering common cybersecurity business scenarios
   • trend-cybertron/Primus-Reasoning: Data regarding reasoning and reflection on cybersecurity tasks for distilled training.

Discover further details in this comprehensive whitepaper released by the Trend Micro AI Lab team on the Cornell University arXiv platform concerning the Primus collection of open-source Datasets for Cybersecurity LLM Training.

3. GitHub (https://github.com/trendmicro/cloud-risk-assessment-agent):
   • Trend Cybertron – Cloud Risk Assessment AI Agent: An open-source endeavor that scans source codes, containers, Kubernetes clusters, and AWS infrastructure to identify CVE vulnerabilities, misconfigurations, and security risks
   • Supplies consolidated risk assessment reports with actionable suggestions and remediation code snippets
   • Featuring an interactive chat interface for investigating security concerns
   • Designed for Gen-AI developers, cloud infrastructure operators, and security teams

Common developer scenarios for Trend Cybertron Open-Source

• Strengthening existing security tools with advanced AI functionalities
• Crafting customized security applications with specialized cybersecurity models
• Training and educating in cybersecurity AI
• Exploring and developing novel security techniques
• Integration into DevSecOps pipelines for automated security evaluations

Trend Micro empowers developers, data scientists, and security experts to draw upon decades of cybersecurity wisdom while participating in the progression of the domain through collaborative innovation in this increasingly intricate digital sphere.

Reasons for Developers to Engage with the Community

If you are engaged in AI application development, there are compelling incentives to interact with Trend Cybertron Open-Source:

1. Reliable Cybersecurity Expertise: Build securely on Trend Micro’s tested security groundwork. Our open-source cybersecurity LLMs, datasets, and agents draw from years of threat intelligence experience and our proven success in safeguarding millions of endpoints globally. When utilizing our AI tools, you are leveraging security insights validated in real-world settings.
2. Access to Specialized Models: Utilize cybersecurity-centric LLMs that have shown substantial enhancements across security benchmarks.
3. Practical Applications: From analyzing threat intelligence to assessing vulnerabilities, detecting malware, and responding to incidents – these tools tackle real-world security issues.
4. Collaborative Innovation: Join a community involved in shaping the future of AI in cybersecurity, with inputs from industry specialists and independent researchers alike.
5. Enterprise-Grade Foundation: Build using technology crafted by a company with 35 years of security know-how and trusted by 20 of the top 25 Fortune Global 500 companies.
6. Influencing Standards: Contribute to a project that is shaping international
7. AI security criteria and regulations.