HiddenLayer secured $50M for its AI-defending cybersecurity tools
HiddenLayer, a startup focused on enhancing the security of AI systems against adversarial attacks, announced that it has secured $50 million in funding. The funding round was co-led by M12 and Moore Strategic Ventures, with participation from IBM, Booz Allen Hamilton, TenEleven, and Capital One. The latest funding brings the startup’s total raised to $56 million. HiddenLayer plans to utilize the funds to support its go-to-market strategies, expand its workforce from 50 employees to 90 by year-end, and invest further in R&D. Chris Sestito, the CEO and co-founder, communicated this via email to TechCrunch.
HiddenLayer is dedicated to cybersecurity and specializes in safeguarding AI from adversarial attacks. The company’s primary focus is on extending detection and response to AI, and they are quickly expanding to meet the market demand for their machine learning security platform. This demand comes from various industries all over the world. HiddenLayer was co-founded in 2019 by Sestito, Jim Ballard, and Tanner Burns. Before this, Sestito led threat research at Cylance, an antivirus startup that BlackBerry later acquired.
The platform offered by HiddenLayer equips data scientists with tools to safeguard their AI models from adversarial attacks, vulnerabilities, and malicious code injections. By closely monitoring the inputs and outputs of AI systems, the platform can assess the integrity of the models before they are deployed.
Many data scientists rely on pre-trained, open-source, or proprietary machine learning models to streamline the testing process and expedite the analysis of complex datasets. While this approach may seem convenient, it can expose organizations to transfer learning attacks resulting from tampered publicly available models. Therefore, exercising caution and taking appropriate measures to protect AI models from potential threats is essential.
To ease customer concerns, HiddenLayer assures that they only have access to the mathematical representations of inputs and outputs, not the proprietary models themselves. This process allows the system to learn what is typical for a unique AI application without explicit instruction. HiddenLayer also contributes to the MITRE ATLAS, a knowledge base of adversarial AI tactics and techniques.
According to Sestito, their platform can protect against all 64 unique attack types listed in the ATLAS, such as IP theft, model extraction, inferencing attacks, model evasion, and data poisoning. While some experts are unsure if HiddenLayer’s methods are groundbreaking, they see the value in packaging knowledge about AI attacks to make it more accessible.
While much research is being conducted on AI security, there must be more public reporting on actual attacks against AI systems. For instance, whether hackers are attempting to breach commercial facial recognition systems is still being determined. However, government agencies are warning about the potential for such attacks.
For example, the National Cyber Security Center in the UK has cautioned about the possibility of threat actors manipulating technology behind large language model chatbots to gain access to confidential information, create offensive content, and cause unintended consequences. Similarly, the US Government’s Office of Science and Technology Policy has released an “AI Bills of Rights,” which recommends pre-deployment testing, risk identification and mitigation, and ongoing monitoring of AI systems to ensure their safety and effectiveness for their intended use.
Many companies are beginning to recognize the importance of protecting their AI models from threats. According to a study by Forrester, most companies currently rely on manual processes to address these threats, while 86% expressed concern about the security of their machine-learning models. Additionally, Gartner reported that 2 in 5 organizations experienced an AI privacy breach or security incident within the past year, with 1 in 4 attacks being malicious.
HiddenLayer, a company specializing in AI security, believes that this threat will only grow as the AI market expands. While other startups offer products designed to make AI systems more robust, HiddenLayer claims to stand out with its AI-driven detection and response approach. The platform has gained traction with partnerships with companies like Databricks and Intel and Fortune 100 customers in various industries, including finance, government, defense, and cybersecurity.
According to HiddenLayer’s CEO, the rapid adoption of AI has left many organizations needing help to implement the necessary processes, people, and controls to protect against risks and attacks. As a result, the demand for their platform is coming from industries across the globe, and they are scaling quickly to meet it.