Fast and accurate fault detection and localization in fiber optic cables is extremely important to ensure the optical network survivability and reliability. Hence there exists a crucial need to develop an automatic and reliable algorithm for real-time optical fiber faults’ detection and diagnosis leveraging the telemetry data obtained by an optical time domain reflectometry (OTDR) instrument. In this paper, we propose a novel data-driven approach based on convolutional neural networks (CNNs) to detect and characterize the fiber reflective faults given noisy simulated OTDR data, whose SNR (signal-to-noise ratio) values vary from 0 dB to 30 dB, incorporating reflective event patterns. In our simulations, we achieved a higher detection capability with low false alarm rate and greater localization accuracy even for low SNR values compared to conventionally employed techniques.