Descriptor-dependent Josephus encryption for cross-script offline signature verification
Abstract
Josephus encryption deterministically rearranges image pixels, but its value for offline signature verification depends on how a descriptor responds to altered spatial neighborhoods. This study evaluates that dependency across scripts under three representation conditions: clear, encrypted, and combined. Four handcrafted representations (local ternary pattern (LTP), oriented basic image features (oBIFs), run-length distribution (RLD), and an LTP–oBIFs fusion) are compared within a writer-independent radial basis function (RBF) support vector machine (SVM) framework. Experiments cover GPDS300, the Dutch and Chinese subsets of SigComp 2011, and the Bengali and Italian subsets of SigWIcomp 2015. Training uses genuine-genuine and genuine-forgery dissimilarity vectors, while testing compares questioned signatures with the mean reference template. The effect of encryption is descriptor- and dataset-dependent. Combined LTP+oBIFs reached 85.06% ACC on GPDS300; clear oBIFs reached 98.08% on Dutch; and clear LTP+oBIFs reached 88.58% on Chinese. For SigWIcomp 2015, the best minimum log-likelihood-ratio costs were 0.0346 for Bengali with combined RLD and 0.0227 for Italian with combined LTP+oBIFs. Josephus encryption can provide complementary evidence in selected settings, but it can substantially degrade orientation-sensitive descriptors, so representation choice should be validated for the target dataset.
Keywords
Cross-script evaluation; Josephus encryption; Local ternary patterns; Offline signature verification; Oriented basic image features; Run-length distributions; Texture descriptors
Full Text:
PDFDOI: https://doi.org/10.11591/eei.v15i6.14147
Refbacks
- There are currently no refbacks.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191
,
e-ISSN: 2302-9285
This journal is published by the
Institute of Advanced Engineering and Science (IAES)
.