User-Friendly Image-based Mnemonic Phrases Management with Stable Diffusion Model for Crypto Wallet
Paper Title: User-Friendly Image-based Mnemonic Phrases Management with Stable Diffusion Model for Crypto Wallet
Journal Name: IEEE Transactions on Dependable and Secure Computing
Abstract: Mnemonic phrases are critical for recovering cryptocurrency wallets and securing digital assets. However, these phrases are often difficult for users to remember and prone to being lost or stored insecurely. This paper proposes a novel, user-friendly approach that combines text-to-image generation using the Stable Diffusion 2.0 model with an image-based mnemonic retrieval process. The system generates multiple images conditioned on a user’s password text and a distraction dataset, allowing users to intuitively recover their mnemonic phrases by selecting the correct images from a set. Security analysis shows that the method provides approximately 7.17 bits of security entropy through a two-step verification process with 12 images per step, effectively balancing usability and security. Experimental results demonstrate that the proposed framework enhances wallet recovery usability without compromising cryptographic security, making it a promising solution for non-custodial cryptocurrency wallets.