ScreenGuard: AI-Powered Image Authenticity Detection for Financial Transaction Fraud Prevention

Navuluri, Satish Reddy (2025) ScreenGuard: AI-Powered Image Authenticity Detection for Financial Transaction Fraud Prevention. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (750kB) | Preview

Abstract

The rising use of online payment platforms has increased the risk associated with fraud using altered screenshots of finance transactions. This has made detecting tampering with a high degree of accuracy extremely important. The paper presents a multimodal deep-learning model that combines the characteristics of RGB images, Error Level Analysis, edge maps, metadata cues and optical character recognition derived results to identify images of forged financial transactions. As the methodology, a labelled dataset was built, modality-specific feature extractors were created, and a multimodal convolutional architecture was built and trained on benchmarked against existing transfer-learning baselines such as EfficientNetV2B0, ResNet50V2 and MobileNetV3Small. The suggested model demonstrated the best validation with an AUC of 1.0, and better accuracy, and the confusion-matrix analysis showed significantly better balance between false positives and false negatives than with image-only models. However, there are still weaknesses in terms of the size of the datasets, OCR sensitivity to domain-specific structure, and capability to withstand increasingly complex adversarial manipulations. This work demonstrates the feasibility of multimodal image-forensic fusion for tampering detection using a publicly available document-tampering dataset. While the experimental evaluation is conducted on generic document images rather than proprietary financial transaction screenshots, the findings should be interpreted as a proof-of-concept indicating potential applicability to fintech fraud-screening workflows. Validation on real financial transaction screenshots remains an important direction for future research.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Mustafa, Raza Ul UNSPECIFIED
Uncontrolled Keywords:	Multimodal image forensics; Financial fraud detection; OCR-based tampering analysis; Error Level Analysis; Deep learning authentication
Subjects:	H Social Sciences > HG Finance Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
Divisions:	School of Computing > Master of Science in Cyber Security
Depositing User:	Ciara O'Brien
Date Deposited:	04 Sep 2026 08:39
Last Modified:	04 Sep 2026 08:39
URI:	https://norma.ncirl.ie/id/eprint/9812

Actions (login required)

View Item