Zero-Trust Federated Learning for Connected Aftermarket Devices

The document is an arXiv cybersecurity research feed covering zero-trust federated learning for connected aftermarket automotive devices, black-box adversarial attacks against OCR vision-language models, security and SOC 2 weaknesses in AI-generated code, structured prompt-injection analysis, identity and verifiable-credential architectures, defenses against reasoning-centric jailbreaks, private risk certification for federated RAG, and conflict attribution under adversarial RAG evidence. The work is primarily research-oriented and does not report a confirmed active exploitation campaign. Notح