Privacy-preserving compression model for efficient IoMT ECG sharing

Abstract Electrocardiogram (ECG) signals are widely used in most remote IoMT systems. Continuous monitoring of patients is required, especially in a pandemic time where doctors recommend telemedicine. This means a massive amount of ECG data is generated, sent to cloud servers, and needs to be shared with legitimate professionals. Therefore, this paper proposes a novel privacy-preserving and efficient technique to reduce the burden on the network while ensuring the privacy of ECG. To ensure efficiency, we use a shallow neural network to learn/remember the ECG shape and represents that in a few neurons. To avoid any loss, the minor residuals between this representation and the original signal is measured and encoded to small footprint using Burrow-Wheeler transform (BWT), followed by move-to-front (MTF) and run-length encoding. To ensure the privacy, only representation neurons are encrypted using a S e s s i o n K e y obtained from the health authority (HA) server along with S e s s i o n I D every-time an ECG signal needs to be transmitted. Hence, health authority alone is able to link that S e s s i o n I D to the patient. Whenever a doctor wants to diagnose an ECG of the patient, HA will share only those two parameters which allow the authorized doctor to see a specific ECG. The model is evaluated using raw ECG data collected from Physio-net. The results obtained are compared and analyzed with the widely used state of art techniques. The results show that the proposed technique outperforms the other techniques by an increase of 50% in size reduction and 60% in transmission time while ensuring the privacy.