Robotics adoption is rapidly growing across several enterprise and consumer deployments. Robots have already become ubiquitous in warehouses and logistics use cases, as well as in manufacturing and industrial use cases; they are also rapidly being adopted across a multitude of applications in healthcare (both within and outside the surgery room), security, construction, cooking, cleaning (robot vacuums), and much more.

However, this rapid growth in adoption has led to many cybersecurity vulnerabilities which, if exploited, can have catastrophic consequences, including data and privacy leakages, industrial sabotage, and even risk of injury – as evidenced by some early examples. In 2017, hackers successfully attacked humanoid domestic robots Alpha2 and NAO, as well as an industrial robot arm sold by Universal Robots. In 2024, hackers took control of robot vacuums across multiple US cities, creating massive security and privacy risks for users. Last year, researchers also identified vulnerabilities in popular humanoid robots manufactured by Unitree. Multiple different research studies have also shown that robots are exposed to a wide variety of cybersecurity risks.

The illusion of security in today’s stack

This high level of risk is due to fundamental architectural flaws in robotics systems – the security stack is currently incomplete. Robotics deployments typically need security that spans across all layers of the technical stack – including physical systems (hardware, sensors, actuators), firmware and embedded systems, operating systems, compute, middleware, networks and communication, data, as well as APIs for connection and integration. Securing these complex, distributed cyber-physical systems requires a cross-disciplinary approach that cannot be transposed from traditional software security methods. However, current security solutions are built across a fragmented, incomplete stack; while security primitives and security tools for robotics exist, they do not protect all these layers simultaneously.

Robot Operating System (ROS) is currently the industry standard for middleware. ROS2, the most adopted version (in 2024, comprising 72% of all ROS downloads), made substantial security improvements over the first version; however, this only scratched the surface. ROS2 introduced Data Distribution System (DDS) based security, which enabled authentication, encryption, and access control. However, security is not enabled by default – it requires manual setup, certificate generation, and policy configurations. Consequently, many deployments run without the security properly enabled, or with misconfigured security. This was also demonstrated by researchers from UC Irvine who, in 2024, found several ROS systems regularly exposed to the internet.

Network security is also quite primitive today across most deployments. Most factories, hospitals, and warehouses using robots tend to have flat networks, with little to no segmentation, and shared WiFi. While this helps with cost, performance, and simplicity of deployment and maintenance, any malicious traffic can easily move laterally across the entire network. Zero-trust network architectures are rarely deployed, leaving a major gap in the security architecture, which should ideally also incorporate continuous authentication and least-privilege communication controls.  

As robot deployments increase, there is a growing need for a fleet layer to orchestrate and coordinate across multiple robots working together. This fleet layer typically uses secured APIs, but this is often accompanied by weak device authentication, shared credentials, and limited policy enforcement. This is worsened by the lack of device-level security, which typically relies on weak or inconsistent identity models, and limited hardware root-of-trust adoption.

Towards a complete robotics security stack

As evidenced above, most robotics deployments have weak security across the layers they are currently defending. In addition, there are also several critical security layers that do not exist in most deployments, but are essential for enterprise-ready mass deployments.

Identity and Access Management (IAM) for robots is a crucial foundational issue that is not yet resolved. While certificates exist in ROS2 and other middleware, they are usually not tied to fleet-wide policies and are rarely rotated. It is critical to set up a unified identity layer that spans the robot, operator, process, and node.

Observability and monitoring tools are also required to continuously track robot operation and flag any anomalous activity that could be indicative of process errors or malicious cyberattacks. Given the complexity of robot deployments and the dynamic nature of the environments they operate in, these need to be physics-aware anomaly detection platforms that allow cross-layer logging, spanning all the sensors, control systems, and networks. These should also enable identification of and debugging non-security issues, such as sensor drift, data drift, or other algorithm-related issues.

Secure deployment and update infrastructures, that protect over-the-air (OTA) updates, are another crucial element. Compromised update packages can propagate across entire fleets and persist below the application layer; trusted update pipelines and supply-chain verification are crucial to mitigate such risks. Standardization of software bill of materials (SBOM) practices can also help – a movement that is being accelerated by the creation of industry standards (such as SPDX and CycloneDX) as well as regulatory mandates such as the EU’s Cyber Resilience Act.

Finally, control-loop security is a critical necessity that is unique to robotics. Robots can be attacked at the control layer through sensor spoofing, control signal injection, and timing manipulation. Real-time anomaly detection systems and control-aware security models are essential to protect against these vulnerabilities.

Conclusion: Securing the future of robotics

As robotics adoption continues to grow, there is a major gap emerging between the scale of deployments and the maturity of the security attack. While there have been no major attacks disclosed yet, this is a growing risk. Mitigating this requires creating a unified and modern security stack that includes several layers – unified IAM rooted in the hardware, middleware with secure-by-default communication, secure supply chains and update pipelines, fleet-wide observability across both physical and digital layers, and real-time physics-aware anomaly detection systems.

Robotics today is like the cloud ecosystem of the early 2010s – scaling rapidly, but riddled with security risks and vulnerabilities. In the cloud world, identity, observability, and other security layers matured into standard infrastructure; several large platforms such as CrowdStrike, Wiz, Datadog, and others emerged to enable this transition. While robotics systems come with much higher levels of complexity and risk, they still depend on inconsistent security tooling that is often bolted-on post deployment. Building and adopting a holistic security stack is crucial to ensuring that robotics deployments can continue scaling in the future.

Vikram Venkat is a principal at Cota Capital, an early-stage venture capital firm focusing on enterprise tech and deep tech, where he covers physical AI, robotics, and cybersecurity. Vikram has an engineering degree from the Indian Institute of Technology Madras, as well as an MBA from The University of Chicago Booth School of Business. He has previously worked at the Boston Consulting Group, Jump Capital, and Rocketship VC.