A Deep Learning-Based Approach for Gender Prediction in Digital Forensics.

In real life, everyone has unique characteristics and abilities that distinguish them from others. Handwriting is one of these characteristics. This feature has an important benefit in many applications, including digital forensics, cybersecurity, and many other fields. The evolution of handwriting systems becomes a necessary and urgent matter due to the development of technology that has increased the use of digital handwriting captured using scanners or electronic pens. In this research, a novel method was applied to enhance the investigation of digital forensics by using handwriting as evidence to determine the gender. The current methods suffer from many limitations related to accuracy and a lack of a uniform technique for evidence verification. This research addresses these limitations by ensuring evidence integrity through SHA-256 cryptographic hashing and performing gender classification using ResNet18-based convolutional neural networks, and thus it can address the limitations of the existing methods by proposing the first dual-purpose forensic framework. Earlier studies focus on either gender prediction or evidence verification, while this study integrates both critical forensic requirements in a unified pipeline. This work is enhancing digital forensics by providing investigators with a reliable, secure, and automated tool for handwriting-based gender identification. The QUWI database was used to evaluate the system, which holds 1,017 handwritten samples in Arabic and English. From the results obtained, we can say that the proposed method has achieved high efficiency in helping criminal investigators by analyzing evidence with high efficiency. The time taken to generate the hash of each document is 0.23 seconds, which means that it exceeds the standard requirement of 0.5 seconds, and this will make the system suitable for real-time forensic applications. Results prove superior performance with 97% gender classification accuracy, significantly outperforming existing methods. [ABSTRACT FROM AUTHOR]