UniqID builds on our published research in hardware-based device fingerprinting — techniques that distinguish real devices (and real users) from bots and spoofed environments using signals rooted in physical hardware, not just browser metadata.
FP-Rowhammer: DRAM-Based Device Fingerprinting
This paper presents a Rowhammer-based fingerprinting approach that leverages DRAM manufacturing variations to produce unique, stable device identifiers — even when software configurations are normalized or obfuscated. Evaluated on 98 DRAM modules, FP-Rowhammer achieves 99.91% fingerprinting accuracy in under five seconds. These hardware-rooted signals are a core building block for distinguishing legitimate devices from automated or spoofed clients.
Read publicationCPU-Print: From Multiplying Matrices to Uniquely Identifying Devices using DVFS
This poster introduces CPU-Print, a DVFS-based device fingerprinting technique that exploits CPU power-draw side channels via controlled workloads (matrix multiplication) to induce frequency-state changes unique to each device. Deployed in the wild with 50,000 traces across 225 devices, CPU-Print achieves up to 88% fingerprinting accuracy — applicable to bot detection, fraud prevention, and stateless device identification from the browser.
Read poster (PDF)UniqID is currently in the R&D phase — we're translating this research into a deployable product. Interested in collaborating or following our progress? Get in touch.