A screenshot from the Password Graveyard website.
Concept
Digital password first appeared on the MIT Compatible Time-Sharing System in the 1960s. Today passwords have become an inseparable part of our lives. Often inspired by personal connections—names of loved ones, meaningful dates, or favorite foods—passwords hold stories and sentiment. However, as the battle between hackers and cybersecurity experts evolves, these meaning-rich passwords are no longer considered secure. With the rise of password managers and passwordless sign-ins, passwords may soon become a technology of the past.
I want to create a space where the “weak” passwords of the past can be remembered. The Password Graveyard website provides a space to explore the semantic connections among outdated passwords. Users can enter a password and discover semantically similar entries from the OWASP common passwords list.
Inspiration
Aram Bartholl’s Forgot Your Password? was a key inspiration for this project, particularly in how it recontextualizes leaked passwords as tangible, public artifacts. By printing thousands of exposed passwords in a physical book, Bartholl transforms something deeply private into an object of collective history. This idea of passwords as fragile yet persistent traces of identity led me to explore their visualization in Password Graveyard.
Technical Challenge
One of the biggest challenges in this project was that using a regular embedding model directly on passwords didn’t produce meaningful semantic connections between passwords. Traditional embeddings focus on semantic relationships between words, but passwords often contain random character combinations, numbers, or inside jokes that don’t have inherent linguistic meaning. To resolve this, I introduced a GPT model into the pipeline, which generates a descriptive sentence based on each password. This added a layer of interpretation, allowing the embedding model to work with richer, contextualized data rather than raw strings.
