AI-powered privacy shields and machine learning approaches for securing digital money transactions: a systematic review

Abstract

Thus, the synergy of artificial intelligence (AI)-based technologies and digital financial transactions require secure anonymized methods while retaining the effectiveness of AI-based fraud-detection. This systematic review investigates stateof-the-art means of enhancing privacy assurance in ML by leveraging innovative schemes to safeguard money transfers in electronic platforms. Many privacy-preserving techniques are available and can be adopted by financial institutions to analyses encrypted data these include homomorphic encryption and federated learning. Employing these methods, AI models can identify fraudulent behavior patterns while at the same time not compromising on the privacy of single transactions. There is an extra level of security or anonymity given x by zero-knowledge proof which allows for the verification of the transactions without disclosing the data behind such transactions. Differential privacy is also used to apply noise on data to ensure that no distinguishing data set is used by the algorithm while ensuring the data is useful for statistical purposes for the ML models used. As much as its integration offers potential in carrying these privacy-shields presents some considerations. Mainly, they improve security and users’ confidence but at the same time introduce computation cost and system intricacy. This review therefore looks at different implementation strategies and hybrid solutions which employ several ideas aimed at maintaining high efficiency of the applied privacy-preserving techniques. Security: Advanced developments in hardware acceleration and algorithms have brought into use these methods nearer to real life applications. It also explores areas of future development including quantum protection of privacy and privacy preserving AI systems. Nonetheless, time and again there are instances where researchers experienced difficulties in the actual implementation such as the approaches may not be scalable, in other words may not well work for large data sets, or that there is need to standardize these models for privacy-preserving AI to be well embraced as it remains one of the most important revolutions by which the safety of financial systems in the digital world can be enhanced. As trading volumes increase and the regulation of how clients’ data is used gets stricter, these technologies will be at the heart of shielding consumer information whilst facilitating enhanced fight against fraud.