HiddenLayer Inc., a startup that helps enterprises shield their artificial intelligence models from hacking attempts, has closed a $50 million funding round. 

Austin-based HiddenLayer announced the Series A round today. It said that Microsoft Corp.’s M12 startup fund and Moore Strategic Ventures led the round. They were joined by Ten Eleven Ventures along with Booz Allen Ventures, IBM Ventures and Capital One Ventures.

Like all software systems, AI models are targeted by cyberattacks. Some hacking campaigns use standard breach techniques such as malicious code injection. Others employ more sophisticated, AI-specific cyberattack methods that can be difficult to detect.

HiddenLayer promises to help companies more easily spot such cyberattacks. It offers a cybersecurity platform built specifically to detect and block breach attempts that target AI models. According to the company, its platform also eases related tasks such as fixing neural network vulnerabilities. 

“We know almost every enterprise in the world is already using AI in multiple forms today, but we also know that no other technology has achieved such widespread adoption without security protections in place,” said HiddenLayer co-founder and Chief Executive Officer Chris Sestito (pictured, center, with co-founders Tanner Burns and James Ballard.)

In some AI-focused cyberattacks, hackers enter malicious input into the targeted system to cause errors. This malicious input often takes the form of seemingly innocuous files that don’t appear malicious to humans, but contain slight modifications capable of confusing a neural network. A hacker might, for example, incorporate such modifications into phishing emails to make them more difficult for AI-powered detection tools to spot. 

Other AI-focused cyberattacks have different objectives. Some aim to steal the training dataset with which a neural network was developed. Other hacking attempts seek to compromise training datasets in a way that will cause the AI models trained on them to malfunction.

According to HiddenLayer, its platform spots cyberattacks by monitoring the inputs entered into a neural network and the output it generates. When the platform spots signs of malicious activity, it notifies administrators. Furthermore, the platform can isolate the hacked neural network and take other steps to minimize the scope of a cyberattack.

Besides detecting breaches, HiddenLayer also promises to help companies prevent cyberattacks from happening in the first place. Its platform includes a tool that can scan an AI model and automatically flag vulnerabilities with the potential to cause a breach. The tool can also spot attempts by hackers to inject malicious code into AI applications.

The third major component of the platform is called Security Audit Reporting. It provides a dashboard that allows users to centrally track the cybersecurity status of their company’s AI models. When it finds vulnerabilities, the tool prioritizes them based on severity to help cybersecurity teams tackle the most urgent issues first.

HiddenLayer says its platform is used by Fortune 100 enterprises in multiple sectors. To advance its revenue growth plans, the company has nearly quadrupled its headcount over the 12 months. It will use the $50 million round announced today to hire an additional 40 staffers by year’s end, build new features and accelerate go-to-market initiatives. 

Photo: HiddenLayer