This paper discusses the design of a model-based fault detection scheme for robust and early detection of runaways in aircraft control surfaces servo-loop. The proposed scheme can be embedded within the structure of in-service monitoring systems as a part of the Flight Control Computer (FCC) software. The final goal is to contribute to improve the performance detection of unanticipated runaway faulty profiles having very different dynamic behaviors, while retaining a perfect robustness. The paper discusses also the tradeoffs between adequacy of the technique and its implementation level, industrial validation process with Engineering support tools, as well as the tuning aspects. The proposed methodology is based on a combined data-driven and system-based approach using a dedicated Kalman filtering. The technique provides an effective method ensuring robustness and good performance (well-defined real-time characteristics and well-defined error rates). Simulation results, using In-flight recorded data sets provided by Airbus, are presented to demonstrate the potential of the developed technique.