No abstract

No abstract

No abstract

The brain structure is unique, hidden and comparatively stable over time than the brain's electrical signals for authentication in high-security areas. Unlike conventional biometric modalities, brain structure needs more security since the theft of the human brain's actual structure is irreversible. Issues such as scalability, uniqueness, robustness against MRI acquisition noise and template security have been insufficiently addressed by the previous methods. The useful brain structures have been segmented using an adaptive segmentation and boundary extraction algorithm. The proposed angular transformation of the original multipronged slices in the coronal, sagittal and horizontal planes, increases the effective surface area and optimizes the structural information in the brain print. Subsequent application of the irreversible layered encryption of the multipronged slices increases the effective number of final brain structural curves in the final brain print. Due to irreversibility, it is impossible to obtain the original brain structure from the final brain print. We have currently tested for 3D brain maps of 209 normal subjects. Template matching has been done through Hausdorff distance among templates. This is also the first method being reported to perform with high accuracy of 99.94% even during noisy MRI acquisition of 10% pixels. The false acceptance rate and false rejection rate in the noisy conditions are 0% and 1.1% respectively. The equal error rate is 8.4%.

Biometric templates must be secured with traceability, immutability, and high-trust capabilities. A variety of system models are proposed by researchers, most of which either utilize blockchains or machine learning for improved security and quality of service (QoS). The augmented sharding model is designed using light weight incremental learning framework, which assists in shard formation and management. Performance evaluation of the proposed model indicates that it is able to achieve high accuracy attack mitigation, along with low block mining delay and high throughput. This performance is compared with various state-of-the-art methods and an improvement of 10% in terms of delay and 14% in terms of throughput is achieved. Further, an attack detection accuracy of 99.3% is obtained for sybil, masquerading, and man in the middle (MITM) attacks. This text further recommends improvement areas which can be further researched for enhancing security and QoS performance of the proposed model.