Competitions
SaTML traditionally includes a Competition Track as part of its program. Participants are invited to engage in selected data science competitions, competing to achieve the highest score on machine learning and security tasks. These tasks are based on well-defined problems and corresponding datasets provided by the competition organizers. The competition results will be presented and discussed during dedicated sessions at the conference.
For this year, the following competitions have been accepted for the conference:
LatentArena tests representation-level safety monitors against black-box attackers. An ensemble of monitors on a disclosed base model covers a hidden subset of harmful-behavior categories. In the discovery phase, participants infer what the monitors watch for. In the evasion phase, they craft inputs that a judge confirms as hazardous but that score below the blocking threshold. One population plays both phases, testing whether identifying a monitor predicts evading it.
Website coming soon
How much warning can ML give before an attack on critical infrastructure takes effect? Can an autonomous agent mitigate attacks without disrupting operations? Ravensport invites participants to defend a realistic critical infrastructure network, running real software on virtualised hosts, as multi-stage attacks unfold. Entrants collect live host and network data with their own tooling and must autonomously detect and mitigate attacks, including methods withheld until final scoring.
Participation and contact
Interested researchers can participate in any of these by following the instructions provided on the competition websites. For more information or specific inquiries, please contact the respective competition organizers directly. For general questions, you can reach us at pcchairs@satml.org.
Please see our Call for Competitions for further details about the competition track.
