Abstract
The Network Time Protocol (NTP) keeps the clocks of Internet-connected devices synchronized, which core applications and protocols depend on for proper functioning, including DNS caches, RPKI, and TLS.
Next to free time servers (NIST, NTP Pool), large cloud providers and OS vendors such as Microsoft, Apple, and Google have begun operating their own time services and setting their billions of systems/devices to use their NTP servers by default.
This creates a level of service centralization around these providers, and yet there has been little scrutiny of these services and their characteristics.
In our paper, we characterize seven large time service providers in terms of replication architecture, client mapping, accuracy and service provided.
We found a large variation in service replication (DNS and anycast), how they map clients to servers, and in their
accuracy and (security) feature support.
This includes violations of best practices, servers clocks being incorrect by up to 50ms on one occasion, the use of outdated NTP versions, and a lack of support for NTS, RPKI, and DNSSEC.
Recording
Video will be added soon.
Speaker
Pascal Huppert
Pascal Huppert is a second-year PhD student in the DACS group at the University of Twente, Netherlands. He performs research on the resilience and security of the NTP (Network Time Protocol) ecosystem.
His most recent paper examines the time servers of the largest tech companies. Pascal hopes to make network time safer and more dependable and welcomes any collaboration on this topic.
Rate this talk
Rating will open: Monday, 26 October 2026 09:00 (+0200).