Survey Report on Forgery Signature Detection System

Abstract

In this review, we present an offline signature forgery detection system utilizing Convolutional Neural Networks (CNN) and Principal Component Analysis (PCA). Handwritten signatures are often forged for fraudulent purposes, necessitating robust detection methods. Our system aims to classify signatures as genuine or forged by extracting key features using CNN, which captures the intricate details of the signature, such as strokes and angles. PCA is applied to reduce the dimensionality of the feature set, ensuring efficient computation without losing critical information. This hybrid approach leverages CNN for its strength in feature extraction and PCA for enhancing the discriminative power of those features. Our model is trained and tested on public datasets, demonstrating significant accuracy improvements over traditional methods, achieving up to 99.7% recognition accuracy. By applying fixed parameter thresholding, our system effectively detects both genuine and random forgeries, minimizing false positives and negatives. This research lays the groundwork for further improvements in forgery detection, proposing a scalable solution for real-world applications.

Introduction

Signature verification is a vital biometric method widely used in legal and financial sectors for identity authentication. However, the rise of signature forgeries—from simple to sophisticated—poses significant challenges. Traditional manual verification is slow and error-prone, prompting the use of automated systems. Recent advances leverage Convolutional Neural Networks (CNNs) for extracting detailed signature features and Principal Component Analysis (PCA) for reducing feature dimensionality to improve efficiency and accuracy.

This paper presents an offline signature forgery detection system combining CNN and PCA, capable not only of classifying signatures as genuine or forged but also identifying forgery types (skilled, unskilled, random). The system preprocesses signature images (noise removal, normalization), extracts hierarchical features via CNN, reduces feature dimensions using PCA, and classifies signatures using cosine similarity with thresholding.

Extensive literature shows that CNNs outperform traditional methods, with hybrid models including PCA enhancing performance by reducing computational complexity while maintaining accuracy. Testing on datasets like CEDAR and SVC2004 demonstrated high accuracy (up to 99.7%), low false positive/negative rates, and effective forgery-type classification.

The system's robustness, efficiency, and scalability make it suitable for practical applications in banking, legal document verification, and digital authentication. Future expansions may include multi-factor authentication and real-time mobile/cloud deployment.

Conclusion

In end, the application of Convolutional Neural Networks(CNN) mixed with Principal Component Analysis (PCA)forForgery signature detection represents a effectivetechnique to enhancing protection in signature verification.The CNN excelsat learning and extracting complicatedfeatures from signaturephotographs, making an allowancefor particular differentiationbetween genuine and solidsignatures. Meanwhile, PCAeffectively reducesdimensionality, which no longer mosteffective hastens processing instances however additionallymitigates theimpact of noise and inappropriate versions insidethe data.This hybrid method has tested a marked improvementindetection accuracy as compared to traditional strategies,making it in particular precious in excessive-stakeenvironments such as banking, legal documentation, andagreement signing. Moreover, the adaptability of CNNsallowsfor continuous improvement as greater facts turninto to be had,paving the manner for even more state-ofthe-art detectionstructures in the destiny.Further studies should consciousnesson refining the model viaincorporating larger and more variousdatasets, exploringadvanced neural community architectures,and investigating realtime implementation possibilities. Thepotential of this approach should revolutionize how weauthenticate signatures, appreciably decreasing fraud andimproving accept as true with in signature based totallytransactions. Overall, this examines highlights the crucialposition of device mastering in preventing forgery and enhancingsecurity features in various domains.

References

[1] Gaurav Yagvalya, Shreya Rawat, Saumya Gupta, Muskan Srivastava, Ashish Shrivastava, “A Novel Method of Fake Signature Detection Using Deep Learning Techniques”, International Journal of Scientific Research in Engineering and Management, Vol.08, Issue 05, May-2024 [2] Navya V K, Abhilasha Sarkar, Aditi Viswanath, Akshita Koul, Amipra Srivastava, “Signature Verification and Forgery Detection System”, International Journal of Creative Research Thoughts, Vol.11, Issue 6, June-2023 [3] Ashish Kumar Srivastava, Dr. Tauseef Ahmad, Jay chand, “Fake Signature Detection Using Neural Networks”, International Journal of Engineering Inventions, Vol.12, Issue 9, Sept. 2023, pp 31-28. [4] Ms. Manjula Subramaniam, Teja E, N Arpith Mathew, “Signature Forgery Detection Using Machine Learning”, International Research Journal of Modernization in Engineering Technology and Science, Volume.04, issue 02, Feb. 2022.

Copyright

Copyright © 2025 T. Arivanantham, Pratiksha Shirsath, Zainab Sayyed, Kiran Potdar, Shubham Sagar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.