DVAIB: A deliberately vulnerable AI bank for practicing prompt injection and AI security attacks

DVAIB (Damn Vulnerable AI Bank) is a deliberately vulnerable AI bank designed for practicing prompt injection and AI security attacks. This platform offers a legal and controlled environment for cybersecurity professionals to hone their skills in AI security. The platform includes three scenarios: Deposit Manipulation, eKYC Document Verification, and Personal Loan. Each scenario targets different aspects of AI security, such as prompt injection, document parsing exploits, and RAG policy disclosure attacks. The technical implications of DVAIB are significant, as it allows professionals to understand and mitigate vulnerabilities in AI systems. By providing a platform for practicing AI security attacks, DVAIB enhances the skills of cybersecurity professionals, which is crucial as AI systems become more prevalent in various industries. From a cybersecurity expert's perspective, DVAIB is a valuable tool for both offensive and defensive security practices. It allows professionals to test their skills in a safe environment, identify potential vulnerabilities in AI systems, and develop strategies to mitigate these risks. For cybersecurity professionals, DVAIB offers actionable intelligence by providing hands-on experience with AI security attacks. This experience can be directly applied to real-world scenarios, helping professionals to better understand and defend against AI-related threats. The practical implications include improved security measures, more effective threat detection, and enhanced incident response capabilities.