The question
Can anomaly detection help identify manipulated measurements—and does the manipulation actually change a downstream decision? This academic project explores that relationship through a motorsport-inspired simulation, with no Formula 1 team affiliation.
My contribution
I completed the project design, implementation and evaluation. This was an individual academic project, using synthetic telemetry and controlled scenarios.
The approach
The prototype compares clean, attacked and defended decisions. It examines anomaly detection alongside decision impact, keeping the simulated setting distinct from real operational telemetry.
Results and limits
The project report describes spikes as easier to detect than gradual bias and distinguishes detection from decision impact. These are reported prototype observations, not independently reproduced benchmarks. Synthetic data, simplified decision rules and controlled attacks limit generalization. No validated production latency or real-world performance improvement is claimed.
Try the signal lab
Explore clean signals, spikes and gradual bias in an illustrative interaction. This is a teaching example, not a replay of the academic experiments.
Open the interactive telemetry labPublic context
The related talk was presented at Rochester Security Summit on October 14, 2026. The event program establishes the speaking appearance; it does not independently validate prototype results.
Read the official RSS session listing ↗Review note: project downloads and detailed technical artifacts are held for content and publication review.