World Congress 2026 Europe Jul 10, 2026 Session details

PET ARENA Season 2.

Sagar Sharma .

Static benchmarks leave privacy systems vulnerable. Now, TikTok and Oblivious have launched a live, peer-to-peer battleground where defenders and attackers co-evolve to battle-test differential privacy architectures.

Pause
Mute Enter Fullscreen
#1 about 2 min

Replacing static privacy benchmarks with adaptive adversarial testing

Peer-to-peer competitions reveal how privacy systems fail against evolving query probes.

#2 about 2 min

Fostering a community of privacy builders and breakers

Data engineers and researchers compete to identify empirical privacy leakages using basic Python and SQL.

#3 about 2 min

Analyzing statistical and logical differencing attack strategies

Prior attackers successfully exploited systems via subgroup inference and reverse-engineered noise mechanisms.

#4 about 2 min

Avoiding common privacy querying and budget management pitfalls

Failing to account for noise floors or exhausting privacy budgets leads to refused queries and lost points.

#5 about 2 min

System architecture orchestrating automated privacy match evaluations

A centralized evaluation hub manages datasets, API gateways, and query proxies to prevent overfitting.

#6 about 1 min

Implementing defense track noise mechanisms and query guardrails

Defenders must balance query suppression and custom noise mechanisms without failing hidden utility thresholds.

#7 about 2 min

Designing inference and record linkage attacks for red teaming

Attackers attempt to extract sensitive attributes and bypass dynamic defenses to reidentify database targets.

#8 about 2 min

Progressing through peer-to-peer cycles and final calibration rounds

Participants complete initial house missions before facing opponent submissions across three competitive cycles.

#9 about 2 min

Calculating leaderboard scores with a dynamic rating system

Continuous match scores scale zero-sum algorithms and adapt K factors based on track participation density.

#10 about 2 min

Preventing match collusion and competing for dual-track rewards

System guardrails involuntarily pair opponents and enforce query limits to ensure equitable grandmaster evaluations.

#11 about 5 min

Building and testing strategies in the browser sandbox environment

A user journey demonstrating how to configure LLM classifiers and execute budgeted attacker notebooks within the interface.

#12 about 3 min

Co-evolving submissions and managing late competition registration limits

Continuous strategy modifications reward participants who acclimate to evolving opponent models early in the cycle.

Matching moments

1:51 min

Exploring inference, extraction, and the adversarial threat landscape

Nura Kawa · WWC 2023

17:05 min

Navigating data privacy boundaries and adversarial model reliability

Alexandra Waldherr · LIVE

4:04 min

Embedding data security and applied ethics into developer education

Daniel Tao +3 · WWC 2024

4:17 min

Overcoming commercial constraints and exploring transparent multiplayer artificial intelligence

Johanna Pirker Johanna Pirker · WWC 2021

1:32 min

Developing fail-safes and defenses through adversarial training

Nura Kawa · WWC 2023

5:55 min

Iterating extensively through evaluations and red teaming

Seppe Housen Seppe Housen · Europe 2026 Virtual

Upcoming sessions on this topic

Open session

World Congress 2026 North America

On the Public Clock: Open-Source Defense When You're Not in the Club

Nicholas Muy

VP Engineering Platform and Security at Scrut.io

Nicholas Muy
Open session

World Congress 2026 North America

RoboCoders: Judgment Day: AI-Assisted Engineering Applied - The Battle of Agents

Baruch Sadogursky, Viktor Gamov

Baruch Sadogursky
Viktor Gamov
Open session

World Congress 2026 North America

SecurePrompt: Building a Pre-Flight Security Layer for Agentic AI

Ravi Sastry Kadali

AI/ML Engineer at General Motors

Ravi Sastry Kadali
Open session

World Congress 2026 North America

Secure-by-Inclusion: Preventing Accessibility Barriers from Becoming Security Vulnerabilities

Radostina (Ina) Tsvetkova

Norwegian Directorate of Labour and Welfare (NAV), Senior Advisor in Digital Accessibility and Inclusive Design

Radostina (Ina) Tsvetkova
Open session

World Congress 2026 North America

Engineering the Pivot: How Creative Strategy Solves the Hard Problems of AI Accuracy and Scale

Shruti Tiwari

AI/ML product manager, Dell

Shruti Tiwari
Open session

World Congress 2026 North America

Responsible AI Architecture with Zero Trust Agents

Ashok Prakash

Staff ML Engineer at Apple

Ashok Prakash