ECG Biometric Recognition Using Adaptive Bloom Filter-Based Template Construction

Biometric recognition, employing physiological or behavioral traits for identity determination, eliminates the need for memorization. Although electrocardiograms (ECGs) show promise as biometric traits, concerns have arisen in existing systems regarding privacy and security due to inadequate template protection. This study introduces Bloom filter-based strategies to generate biometric templates suitable for ECG biometric systems, whether in identification or verification mode. The incorporation of nonlinear transformations during template construction complicates the conversion process, making the reconstruction of ECG heartbeat segments from their templates challenging, addressing existing privacy and security concerns. Identity matches are further confirmed using an interquartile range-based method, enhancing recognition accuracy amid illegitimate access attempts and inter-beat variation. Experimental results demonstrated that the proposed schemes achieved mean equal error rates of 7.9% in identification mode and 1.3% in verification mode.