🤖 machine learning

HEART: A High-Efficiency Adaptive Real-Time Telemonitoring Framework for Secure Electrocardiogram Signal Transmission Using Chaotic Encryption

This paper presents HEART, a secure real-time telemonitoring framework that utilizes a learnable key generator derived from individual ECG signals to dynamically configure chaotic encryption, thereby ensuring confidential transmission and near-lossless reconstruction of patient data while eliminating traditional key exchange vulnerabilities.

Original authors: Beyazıt Bestami Yuksel

Published 2026-05-12

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Beyazıt Bestami Yuksel

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a very sensitive, private diary written in a special code that only you and your doctor can read. Now, imagine that every time you write a new page, the ink, the paper texture, and even the alphabet you use change instantly based on your own heartbeat. That is essentially what this paper, HEART, is about.

Here is a simple breakdown of how this system works, using everyday analogies:

1. The Problem: Sending a Secret Message

In modern telemedicine, patients wear devices that record their heartbeats (ECG signals) and send them to doctors over the internet. The problem is that sending this data is like shouting your medical secrets across a crowded room; anyone listening could steal or tamper with the information. Traditional "locks" (encryption) are often too heavy and slow for these devices, causing delays that could be dangerous in an emergency.

2. The Solution: A "Living" Lock

The authors created a system called HEART that acts like a living, breathing lock. Instead of using a static password (like "123456") that stays the same forever, this system creates a unique, one-time password for every single chunk of heart data it sends.

The Biometric Key: Imagine your heart has a unique "fingerprint" in its rhythm. The system looks at a tiny 0.6-second slice of your heartbeat, measures its average speed and how much it wiggles (statistics), and uses those numbers to generate a secret code.
The Chaotic Mixer: This code controls a mathematical recipe called a "logistic map." Think of this as a high-speed blender. It takes your heart data and mixes it up so thoroughly that it looks like static noise on a TV screen. It does two things:
1. Permutation (Shuffling): It rearranges the order of the data points, like shuffling a deck of cards.
2. XOR Diffusion (Scrambling): It flips the bits of the data, like changing every "yes" to "no" and vice versa, based on a chaotic pattern.

3. The "Smart" Assistant (Machine Learning)

The paper introduces a clever twist: a Machine Learning (ML) assistant.

The Direct Way: Usually, the system calculates the secret code directly from the heartbeat.
The Smart Way: Sometimes, heartbeats are noisy (like trying to hear a whisper in a storm). The ML assistant acts like a seasoned detective who has studied thousands of heartbeats. It predicts what the secret code should be, even if the signal is a bit messy. This makes the lock more stable and harder to break when the data isn't perfect.

4. The Journey: From Patient to Doctor

Here is the step-by-step flow described in the paper:

Capture: A portable device grabs your heartbeat.
Lock: The system instantly turns that heartbeat into "noise" using the unique key generated just for that moment.
Send: This scrambled noise is sent to the cloud. Even if a hacker steals it, it looks like random garbage.
Unlock: The doctor's computer has the matching key (generated the same way or predicted by the AI). It reverses the process, turning the noise back into a clear heartbeat picture.
Diagnose: While this is happening, a computer brain (a neural network) looks at the data to spot heart diseases, all while the data remains secure.

5. How Strong is the Lock?

The authors tested this system rigorously, like a security guard testing a vault:

Randomness: The scrambled data looks perfectly random (like white noise), with no patterns left behind.
Sensitivity: If you change the key by even a tiny fraction (like changing one letter in a password), the whole message becomes unrecognizable.
Speed: The locking and unlocking happen incredibly fast (in milliseconds), so there is no delay in real-time monitoring.
Fidelity: When the doctor unlocks the message, the heart signal looks almost exactly the same as the original (99.99% accurate), preserving the details needed for diagnosis.

6. The Catch (What the Paper Admits)

While the system is excellent at keeping the data secret and fast, the paper notes a small imperfection: when the data is unlocked, some very specific, tiny details of the heart's shape (like the exact height of certain waves) might be slightly off. It's like a photocopy that is 99% perfect, but the very faintest lines might be slightly blurry. For general monitoring, this is fine, but for ultra-precise medical measurements, the authors say they need to tweak the "unblurring" process in the future.

Summary

The HEART framework is a high-speed, secure way to send heart data. It uses your own heartbeat to create a unique, unbreakable code for every second of data, scrambles it so thoroughly that it looks like noise, and then unscrambles it instantly for the doctor. It combines math, biology, and artificial intelligence to keep your medical secrets safe without slowing down the care you receive.

Technical Summary: HEART Framework for Secure Real-Time ECG Transmission

Problem Statement
The increasing reliance on telemedicine and remote patient monitoring has created a critical need for securing sensitive biomedical data, specifically electrocardiogram (ECG) signals. While traditional encryption methods ensure privacy, they often introduce significant computational overhead and latency, which can hinder real-time diagnosis and response. Furthermore, static key exchange mechanisms lack the dynamism required for continuous, personalized patient monitoring. The challenge lies in developing a system that offers robust security, low latency for real-time processing, and the ability to preserve diagnostic fidelity while adhering to strict regulatory standards (e.g., GDPR, HIPAA).

Methodology
The paper proposes HEART (High-Efficiency Adaptive Real-Time Telemonitoring Framework), a system designed for the secure acquisition, transmission, and storage of ECG signals using a hybrid chaotic encryption approach. The methodology operates on 300-sample segments (approx. 0.6 seconds) acquired at 500 Hz via a 3-lead preamplifier.

Biometric Key Generation:
The system employs a Learnable Key Generator (LKG) that derives encryption parameters directly from the patient's own ECG signal. For each segment, the mean ( $\mu$ ) and standard deviation ( $\sigma$ ) are calculated to dynamically determine the control parameter ( $r$ ) and initial condition ( $x_0$ ) for a logistic map:
- $r = 3.6 + (\sigma \mod 0.4)$
- $x_0 = 0.1 + (\mu \mod 0.8)$
  This ensures that the encryption key is unique to the specific biometric segment.
Chaotic Encryption Scheme:
The encryption process utilizes a logistic map to generate a chaotic sequence, which drives two distinct phases:
- Permutation: The sequence generates an index to shuffle the order of the signal samples.
- Diffusion (XOR Masking): The sequence is converted into an XOR mask to alter the actual values of the signal.
  This combination of permutation and XOR diffusion addresses the limitations of permutation-only schemes by actively randomizing data values.
Machine Learning Integration:
To enhance robustness against noise and variability, the system incorporates a Multi-Layer Perceptron (MLP) neural network. This model is trained to predict the chaotic parameters ( $r, x_0$ ) from normalized ECG features. This allows for a "predictive" encryption mode where keys are inferred rather than strictly derived, offering a secondary layer of adaptability.
Decryption and Diagnosis:
Decryption reverses the process using the inverse permutation and XOR operation. The framework operates in an end-to-end pipeline where encrypted data is stored in the cloud, while decrypted data is simultaneously processed by a Convolutional Neural Network (CNN) for real-time disease detection and visualization.

Key Contributions

Enhanced Chaotic Encryption Architecture: The authors propose a novel system combining chaotic permutation with an XOR diffusion layer, specifically tailored for real-time ECG signals to ensure both confusion and diffusion.
Biometric-Based Dynamic Keying: Unlike static or session-level keys, the system generates per-segment keys derived from the ECG signal itself. This is further stabilized by an ML model, creating a personalized and adaptive cryptographic mechanism.
Comprehensive Security Validation: The study provides an extensive suite of security analyses, including Shannon entropy, NIST SP 800-22 frequency testing, correlation/autocorrelation disruption, histogram statistics, and sensitivity tests (both key and plaintext).
Reproducibility and Open Source: The authors have released all source code, test scripts, and processed datasets to ensure independent benchmarking and reproducibility.

Experimental Results
The system was evaluated using data from 1,000 different patients. Key performance metrics include:

Security Metrics:
- Entropy: Encrypted signals achieved Shannon entropy values between 7.6 and 7.8 bits (approaching the ideal 8.0 for 8-bit data).
- Correlation: Adjacent pixel correlation was reduced to near zero ( $|r| < 0.02$ ), effectively removing statistical dependencies.
- NIST Testing: Encrypted sequences passed the NIST 800-22 Monobit frequency test (p-values > 0.01), whereas original signals failed.
- Sensitivity: The system demonstrated high sensitivity to key changes (avalanche effect) and plaintext modifications.
- Key Space: While the global $(r, x_0)$ space is estimated at ~12.94 bits, the authors clarify that security relies on the per-segment refresh, salted parameterization, and diffusion, rather than the raw key space size alone. Min-entropy accounting was used to provide distributional lower-bounds against exhaustive search.
Performance and Fidelity:
- Latency: Average encryption time was 0.0012 seconds (1.2 ms), and decryption was 0.0001 seconds (0.1 ms), confirming suitability for real-time applications.
- Reconstruction Quality: The system achieved near-lossless decryption with a Mean Squared Error (MSE) of $\approx 5 \times 10^{-6}$ , Peak Signal-to-Noise Ratio (PSNR) > 52 dB, and Mean Absolute Error (MAE) $\approx 0.002$ .
- Robustness: The system maintained low error rates under noise and occlusion attacks.

Significance and Claims
The paper claims that the HEART framework successfully bridges the gap between high-security requirements and the low-latency demands of real-time telemedicine. By integrating biometric key generation with chaotic encryption and machine learning, the system offers a scalable solution that eliminates traditional key exchange overheads.

The authors modestly note that while global reconstruction fidelity is high and sufficient for general monitoring and visualization, specific morphological features (such as QRS amplitude and PR intervals) showed measurable deviations after decryption. Consequently, they assert that while the system is effective for secure transmission and broad diagnostic insights, further refinement is necessary to guarantee the preservation of all clinically sensitive markers (e.g., ST-segment changes) for definitive diagnosis. The study positions this architecture as a promising direction for privacy-preserving biomedical signal processing, with potential extensions to other signals like EEG and EMG.