In this age of digital transactions, online payment fraud has grown to be a serious threat. The complexity of fraud schemes that target both individuals and businesses is increasing along with the global growth of e-commerce and digital banking. The increasing risk of financial fraud has become a major problem globally where wireless communication is essential for sending enormous amounts of data when protecting against interference. In this paper, Advanced Detection of Online Payment Fraud in Digital Transactions through Optimized Attention Augmented Residual Convolutional Neural Network (DOPF-DT-AARCNN) is proposed. Here, the input data is collected through IEEE-CIS Fraud Detection Dataset. Then, the input data is fed into preprocessing stage. In preprocessing, Sub Aperture Keystone Transform Matched Filtering is employed for handling missing data, removing duplicate records and standardizing data scaling. Then, the pre-processed data is supplied to the Feature-Weighted Synthetic Minority Over-sampling Technique for data balancing. The balanced data is given into High-Order Time-Reassigned Synchrosqueezing Transform for extracting the transaction features such as Transaction Amount, Transaction Type, Time, and account balance after transaction. The extracted features are given to the Attention Augmented Residual Convolutional Neural Network (AARCNN) for security and detecting Online Payment Fraud in Digital Transactions as fraud and no fraud. Then, the Dynamic Hunting Leadership Optimization is used to enhance the parameters of AARCNN. The DOPF-DT-AARCNN attains 6.28%, 8.78% and 5.29% higher accuracy and 9.45%, 6.22% and 7.28% higher precision when compared to the existing techniques: Online Payment Fraud Detection utilizing Machine Learning Techniques, A method for card fraud detection under Catboost and deep neural network and A Graph Neural Network along Reinforcement Learning for Adaptive Financial Fraud Detection respectively.