Fujitsu introduced a new multi-AI agent security technology as part of its AI initiative, Fujitsu Kozuchi. This system leverages a coordinated network of specialized AI agents developed to simulate cyberattacks, protection strategies and formulate business continuity measures against evolving threats.
The architecture of this system is based on three pivotal technologies: the Security AI agent, multi-AI agent collaboration, and generative AI security enhancement.
Fujitsu will commence field trials for the generative AI security enhancement technology in partnership with Cohere Inc., starting in December 2024.
Moreover, Fujitsu intends to release portions of the multi-AI agent collaboration framework as open-source software (OSS) on OpenHands—an AI agent platform developed with significant collaboration from Carnegie Mellon University—in January 2025.
Following this, a comprehensive rollout of the technologies is planned globally in March 2025 on a trial basis.
Security AI Agent
This innovative technology incorporates three foundational AI agents for proactive security measures, utilizing methodologies co-developed with the Ben-Gurion University of the Negev, a leader in cybersecurity-related research.
Attack AI Agent: Employs a tactics, techniques, and procedures (TTP) analogy engine to propose effective attack simulations against target systems, addressing new cyber threat vectors.
Defense AI Agent: It generates countermeasures tailored to the organization’s risk profile.
Test AI Agent: Automatically constructs a verification virtual environment (cyber twin) modeled on the production system to evaluate the impact of proposed countermeasures.
This agent facilitates reciprocal attack-defense simulations with the other agents, fostering proactive defenses against emerging vulnerabilities. The collaborative, adversarial learning framework of these AI agents is designed to address complex, previously uncharted security threats.
Multi-AI Agent Collaboration
This feature enables seamless interaction among AI agents, with broader applicability beyond security domains. Addressing challenges regarding data confidentiality, privacy, and adherence to organizational protocols, Fujitsu has launched a secure agent gateway coordination function within the infrastructure of each organization’s AI agents. It allows for autonomous policy control during collaboration, enabling effective AI agent interaction without requiring explicit input from developers.
Generative AI Security Enhancement
In anticipation of new threats targeting generative AI, Fujitsu has collaborated with Ben-Gurion University to create an LLM vulnerability scanner capable of comprehensive security assessments and an LLM guardrail mechanism to defend against cyberattacks.
The vulnerability scanner evaluates over 3,500 contemporary vulnerabilities present in GenAI, including proprietary checks for generative code vulnerabilities. Additionally, the vulnerability explanation functionality brings clarity to vulnerability assessments, making them accessible to non-cybersecurity experts.
Associate Professor Graham Neubig of Carnegie Mellon University commented, “The development of AI agents is rapidly advancing worldwide. OpenHands, an open-source project strongly supported by Carnegie Mellon University, facilitates the development of diverse AI agent services, functionalities, and technologies by incorporating external expertise. Combining Fujitsu’s multi-AI agent collaboration technology with OpenHands could significantly accelerate the development of cross-organizational multi-AI agent applications, which would be a remarkable achievement.”
Seraphina Goldfarb-Tarrant, Head of Safety, Cohere Inc, stated, “Secure systems are mission-critical for enterprises to adopt and benefit from AI development. We’re excited that our ongoing work with Fujitsu enables Japanese companies to leverage AI with our best-in-class multilingual models and security enhancement technology.”
The integration of Fujitsu’s multi-AI agent collaboration technology with OpenHands could significantly enhance the landscape of AI agent solutions.
Also Read: GenAI is a double-edged sword for Defence and Offense in cybersecurity
