Website Fingerprinting (WF) shows that encrypting content is not enough to protect browsing privacy. A passive observer can still infer visited sites by analyzing traffic patterns such as packet sizes, packet timing, direction, and burst structure. Recent deep learning methods make this inference easier and more robust, as they can learn directly from sequences of sizes and inter arrival times even under noisy network conditions. With the widespread use of QUIC and HTTP/3, these patterns remain observable at the network level and therefore continue to be a relevant side channel. This thesis presents a client-only countermeasure for QUIC and HTTP/3 traffic implemented with eBPF.