<p>Accurate time synchronization is a key requirement in today’s digital world. Considering the distributed deployments, it is important that all the devices and systems of digital infrastructure like servers, computers, network devices, Internet of Things (IoT) devices, etc., remain synchronized and work in unison. The Network Time Protocol (NTP) plays a crucial role in making this possible, ensuring consistent and reliable time across devices over the network/Internet. To handle a high number of time requests and maintain dependable service, organizations often run multiple NTP servers. In such cases, it becomes essential to make sure all these servers are precisely synchronized and providing correct time to the end users. There are many public NTP servers accessible freely on the Internet; however, ascertaining timing accuracy and avoiding any anomalous behaviours at the user end is a critical task. Anomaly detection techniques enable early identification of deviations from usual behaviour, allowing proactive mitigation before issues escalate. This paper presents a simple yet effective prototype for detecting anomalies in NTP servers’ data. The approach is based on analyzing NTP packets received at a remote system; therefore, no special access or privileges are required on the NTP servers under observation, beyond standard NTP message exchange. A Python script is developed to collect key metrics from NTP packets. The system pre-processes the NTP packet exchange data and uses statistical analysis methods and machine learning technique (Isolation Forest) to spot any abnormal patterns. The work shows promising results, making it a useful tool for improving service reliability. The work may be enhanced with implementation of advanced machine learning methods and consideration of additional parameters.</p>