Eight months after its previous nine-figure round, Upwind Security Inc. has raised $300 million in additional funding.

Returning bakers Bessemer and TCV led the Series C deal. CTech reported on Wednesday that the firms were joined by Salesforce Ventures, Greylock, Craft Ventures and several others. Upwind is now valued at $3.8 billion, $2 billion more than at the start of the year.

One of the main challenges involved in securing cloud environments is that they change on a regular basis. Developers frequently add and remove workloads, which makes it difficult to ensure that every software asset adheres to data protection policies. The same issue applies to instances and user accounts.

Upwind sells a platform that can automatically map out the assets in a cloud environment. It refreshes the data every 30 seconds to account for configuration changes.

One of the most effective tools for collecting data about cloud environments is the Linux operating system. Its kernel, which contains its most important components, can gather a significant amount of telemetry about user workloads. The challenge is that putting data from the kernel to use requires highly complicated, error-probe operating system changes.

Upwind overcame that challenge using a Linux feature called eBPF. It makes it possible to adapt the operating system for data collection without error-prone changes. The feature deploys observability code in an isolated sandbox, which prevents bugs from finding their way into sensitive kernel modules.

The service collects data about instances, encryption keys, configuration scripts and a range of other cloud assets. It also maps out how those assets interact with one another. For example, administrators can use the software to determine which documents a data visualization application can access.

Several of Upwind’s observability features focus on monitoring artificial intelligence models. The platform can automatically generate so-called AI-BOMs, files that list AI applications’ bills of materials, or components. Such lists make it easier for administrators to spot risky software.

Upwind uses AI to scan the cloud assets that it finds for weak points. The platform compares each workload’s configuration against cybersecurity best practices and regulatory requirements. Additionally, it launches simulated cyberattacks against important assets. For example, the platform might attempt to open a connection to a cloud instance that contains a SQL database.

Vulnerabilities don’t always have a realistic chance of leading to a breach. According to Upwind, its platform filters low-risk flaws using an AI agent called Red. Another agent known as Blue spots hackling attempts. 

It’s unclear how Upwind plans to use the proceeds from its round. Over the past month, the company has rolled out integrations with about a half-dozen third-party cloud services. That suggests its current growth roadmap places an emphasis on increasing the number of systems it can secure for customers. 

Photo: Upwind