Modern Smart Health Systems (SHS) involve the concept of connected personal medical devices. These devices significantly improve the patient's lifestyle as they permit remote monitoring and transmission of health data (i.e., telemedicine), lowering the treatment costs for both the patient and the healthcare providers. Although specific SHS communication standards (i.e., ISO/IEEE 11073) enable real-time plug-and-play interoperability and communication between different personal medical devices, they do not specify any features for secure communications. In this paper, we demonstrate how personal medical device communication is indeed vulnerable to different cyber attacks. Specifically, we show how an external attacker can hook into the personal medical device's communication and eavesdrop the sensitive health data traffic, and implement manin-the-middle, replay, false data injection, and denial-of-service attacks. Furthermore, we also propose an Intrusion Detection System (IDS), HEKA, to monitor personal medical device traffic and detect attacks on them. HEKA passively hooks into the personal medical traffic generated by medical devices to learn the contiguous sequence of packets information from the captured traffic and detects irregular traffic-flow patterns using an n-grambased approach and different machine learning techniques. We implemented HEKA in a testbed consisting of eight off-the-shelf personal medical devices and evaluated its performance against four different attacks. Our extensive evaluation shows that HEKA can effectively detect different attacks on personal medical devices with an accuracy of 98.4% and Fl-score of 98%.