Hacker Rock & Roll

The talk looks back at roughly twenty years of ShmooCon, covering about 800 talks. The speakers found the records scattered across several sources rather than in one place, so they gathered, cleaned and merged them into a single spreadsheet. The project was inspired by an art piece that charted music genres over time, which led to a similar view of the conference's themes. Fire talks were left out because the early years did not record them consistently.

To organize the material, the team built a taxonomy from the bottom up, letting the talks shape the categories, and ended with 45 top-level groups. Each talk was labeled from its title and abstract, then from the video or news coverage when those were not enough, using a majority-of-content rule for talks that span several areas. Exploitation was the largest category, with law and policy, defense, detection and intelligence also prominent. Trends shifted over time: deception topics first appeared around 2010, and new subtopics tended to follow new technologies. The graphs were drawn with Graphviz, using a Python parser that converts a tab-separated file into valid graph code.

The speakers released the dataset, the parsing code and high-resolution visualizations for download. They also noted that the corpus reflects accepted talks rather than all submissions, so it carries selection bias. Their takeaways were that volunteer archivists and archive sites deserve support, and that studying a conference's full history is a useful way to learn the field.