Today, we’re saying Sec-Gemini v1, a brand new experimental AI mannequin targeted on advancing cybersecurity AI frontiers.
As outlined a yr in the past, defenders face the daunting process of securing in opposition to all cyber threats, whereas attackers must efficiently discover and exploit solely a single vulnerability. This elementary asymmetry has made securing methods extraordinarily troublesome, time consuming and error susceptible. AI-powered cybersecurity workflows have the potential to assist shift the stability again to the defenders by pressure multiplying cybersecurity professionals like by no means earlier than.
Effectively powering SecOps workflows requires state-of-the-art reasoning capabilities and in depth present cybersecurity information. Sec-Gemini v1 achieves this by combining Gemini’s superior capabilities with close to real-time cybersecurity information and tooling. This mixture permits it to attain superior efficiency on key cybersecurity workflows, together with incident root trigger evaluation, menace evaluation, and vulnerability affect understanding.
We firmly consider that efficiently pushing AI cybersecurity frontiers to decisively tilt the stability in favor of the defenders requires a powerful collaboration throughout the cybersecurity neighborhood. This is why we’re making Sec-Gemini v1 freely obtainable to pick out organizations, establishments, professionals, and NGOs for analysis functions.
Sec-Gemini v1 outperforms different fashions on key cybersecurity benchmarks because of its superior integration of Google Threat Intelligence (GTI), OSV, and different key knowledge sources. Sec-Gemini v1 outperforms different fashions on CTI-MCQ, a number one menace intelligence benchmark, by at the least 11% (See Figure 1). It additionally outperforms different fashions by at the least 10.5% on the CTI-Root Cause Mapping benchmark (See Figure 2):
Figure 1: Sec-Gemini v1 outperforms different fashions on the CTI-MCQ Cybersecurity Threat Intelligence benchmark.
Figure 2: Sec-Gemini v1 has outperformed different fashions in a Cybersecurity Threat Intelligence-Root Cause Mapping (CTI-RCM) benchmark that evaluates an LLM’s potential to know the nuances of vulnerability descriptions, establish vulnerabilities underlying root causes, and precisely classify them in accordance with the CWE taxonomy.
Below is an instance of the comprehensiveness of Sec-Gemini v1’s solutions in response to key cybersecurity questions. First, Sec-Gemini v1 is ready to decide that Salt Typhoon is a menace actor (not all fashions do) and supplies a complete description of that menace actor, because of its deep integration with Mandiant Threat intelligence knowledge.
Next, in response to a query concerning the vulnerabilities within the Salt Typhoon description, Sec-Gemini v1 outputs not solely vulnerability particulars (because of its integration with OSV knowledge, the open-source vulnerabilities database operated by Google), but additionally contextualizes the vulnerabilities with respect to menace actors (utilizing Mandiant knowledge). With Sec-Gemini v1, analysts can perceive the chance and menace profile related to particular vulnerabilities quicker.
If you have an interest in collaborating with us on advancing the AI cybersecurity frontier, please request early entry to Sec-Gemini v1 through this way.