This study examines privacy risks in collaborative robotics, focusing on the potential for traffic analysis in encrypted robot communications. While previous research has explored low-level command recovery, our work investigates high-level motion recovery from command message sequences. We evaluate the efficacy of traditional website fingerprinting techniques (k-FP, KNN, and CUMUL) and their limitations in accurately identifying robotic actions due to their inability to capture detailed temporal relationships. To address this, we introduce a traffic classification approach using signal processing techniques, demonstrating high accuracy in action identification and highlighting the vulnerability of encrypted communications to privacy breaches. Additionally, we explore defenses such as packet padding and timing manipulation, revealing the challenges in balancing traffic analysis resistance with network efficiency. Our findings emphasize the need for continued development of practical defenses in robotic privacy and security.