Quick UDP Internet Connection (QUIC) is an emerging end-to-end encrypted, transport-layer protocol, which has been increasingly adopted by popular web services to improve communication security and quality of experience (QoE) towards end-users. However, this tendency makes the traffic analysis more challenging, given the limited information in the QUIC packet header and full encryption on the payload. To address this challenge, a novel rule-based approach is proposed to estimate the application-level traffic attributes without decrypting QUIC packets. Based on the size, timing, and direction information, our proposed algorithm analyzes the associated network traffic to infer the identity of each HTTP request and response pair, as well as the multiplexing feature in each QUIC connection. The inferred HTTP attributes can be used to evaluate the QoE of application-layer services and identify the service categories for traffic classification in the encrypted QUIC connections.