Toward Forensic-Friendly AI: Integrating Blockchain with Federated Learning to Enhance AI Trustworthiness

Toward Forensic-Friendly AI: Integrating Blockchain with Federated Learning to Enhance AI Trustworthiness

연구 분야: Safety

학회: International Conference on Digital Forensics and Cyber Crime

Artificial intelligence (AI) systems have increasingly become integral to various industries, such as healthcare, finance, and manufacturing. This integration presents complicated security issues affecting system performance and decisions. AI-related incidents may target the system itself, making it a victim, or the AI system itself, causing harm to others; in both cases, this will result in technical, ethical, legal, and societal consequences. The unique properties of AI systems necessitate the development of specialized AI forensics. Traditional digital forensics domains, including computer, network, database, and mobile forensics, prove demonstrably inadequate when confronted with the opaque nature of AI systems, often referred to as black boxes. This paper proposes a novel forensics-friendly schema to increase forensic readiness within AI systems. In this paper, Federated Learning (FL) is used as a case study to illustrate the application of this schema. The proposed approach integrates blockchain technology with the federated learning environment, making it more forensics-friendly and leveraging immutability, integrity, and transparency features. This enables the integration of complete logs with blockchain, promoting a forensically equipped environment.