Handling high level of uncertainty in forensic signature examination

Abstract

ABSTRACT Signature verification often entails ambiguous scenarios where a signature may closely resemble multiple reference samples or differ due to inherent handwriting variations. In such cases, a neutrosophic engine can enhance the verification process by addressing the uncertainties and ambiguities present in signatures. This is achieved by representing the features and similarity scores produced by the matching algorithm as neutrosophic sets, thereby capturing the degrees of membership, non-membership, and indeterminacy. However, type-1 neutrosophic logic, which assigns fixed values to these membership functions, proves inadequate in capturing the varying degrees of uncertainty in signature characteristics. This limitation stems from its inability to adjust to different levels of uncertainty. To address these shortcomings, the proposed forensic signature examination approach explores the utilization of type-2 neutrosophic logic. This approach offers greater flexibility and granularity in handling ambiguity, indeterminacy, and uncertainty, thereby enhancing the accuracy of signature verification systems. Type-2 neutrosophic logic facilitates more precise decision-making by allowing for the evaluation of multiple sources of ambiguity and conflicting information. Experimental results demonstrate the advantages of employing a type-2 neutrosophic engine for signature verification, particularly in its superior capability to manage uncertainty and variability when compared to type-1 neutrosophic logic. This advancement leads to more accurate False Rejection Rate (FRR) and False Acceptance Rate (FAR) outcomes. Comparative analysis using a benchmark dataset of handwritten signatures reveals that the type-2 neutrosophic similarity measure achieves an accuracy rate of 98%, outperforming the 95% accuracy rate associated with type-1 neutrosophic logic.