Microsoft Introduces Project Perception as a New Agentic Security System to Protect Against AI-Powered Cyberattacks

Microsoft has introduced a new cyber stack that brings together multiple AI agents to work together and offer enhanced security against cyberattacks. Dubbed Project Perception, the cyber stack primarily aims to protect organisations against AI-powered cyber threats by using multiple AI agents that continuously learn system defences, bringing together signals, context, AI models, and specialised agents. The new Project Perception enables three classes of AI agents to work in coordination to identify and investigate threats from AI-powered attacks, while also taking corrective action against them. The cyber stack itself comprises multiple layers, including the specialised agents and AI models, with specified capabilities.

Microsoft Unveils Project Perception

In a blog post on Monday, the Redmond-based tech giant revealed details about its new Project Perception, which aims to provide enhanced protection against AI-powered cyber threats by using specialised AI models and agents to defend against them. The cyber stack “brings together signals, context, models and specialised agents”, which can reason, prioritise, and autonomously take action against such threats.

The Project Perception coordinates three classes of specialised agents, namely Red, Blue, and Green team agents. The Red team agents are built to identify ways through which a bad actor can exploit to compromise an organisation's tech architecture. Meanwhile, the Blue team agents are capable of investigating, contextual reasoning, and determining potentially “meaningful risks”. Lastly, the Green team agents work to remediate by taking corrective actions to strengthen defences.

Microsoft's new cyber stack comprises multiple layers
Photo Credit: Microsoft

Microsoft says that the three classes of AI agents form a “closed-loop system” that can “continuously” discover, evaluate, and enhance an organisation's digital security architecture. The Redmond company has announced that Project Perception will be available for public preview on August 3. Further, the tech giant highlighted that apart from AI agents, an improved defence system also requires a revamped cyber stack.

The tech giant elaborates that its new cyber stack comprises multiple layers, starting with signals and sensors that provide visibility across endpoints, identities, data, clouds, and apps. On top of it lies the context layer, which is capable of continuously enriching intelligence to give agents operational context. A layer of multiple AI models, which can reason over threats with expertise, sits on top of it, along with orchestrators, specialised agents, and actuators, which determine how an agent's decision translates into real-world scenarios.

To highlight the need for an evolved cyber stack, like Project Perception, Microsoft pointed out that the speed at which bad actors can deploy attacks has increased with the adoption of AI, while also becoming increasingly cost-efficient. “Attackers can generate exploits faster, scale campaigns further, and operate with unprecedented efficiency. The approaches built for a world of human actors cannot keep pace with a world of AI, agents, and machine-speed attacks,” the company said.