Requiem for a drone: a machine-learning based framework for stealthy attacks against unmanned autonomous vehicles

TL;DR

Requiem is a machine-learning based framework that stealthily exploits modeling uncertainties in autonomous vehicles, causing mission deviations without detection by onboard systems.

Contribution

It introduces a software-only, blackbox attack framework that leverages deep learning models to manipulate sensor data and evade detection in UAVs.

Findings

Successfully causes significant deviations in UAV missions.

Remains undetected by onboard anomaly detectors.

Generalizes across different sensors and attack types.

Abstract

There is a space of uncertainty in the modeling of vehicular dynamics of autonomous systems due to noise in sensor readings, environmental factors or modeling errors. We present Requiem, a software-only, blackbox approach that exploits this space in a stealthy manner causing target systems, e.g., unmanned aerial vehicles (UAVs), to significantly deviate from their mission parameters. Our system achieves this by modifying sensor values, all while avoiding detection by onboard anomaly detectors (hence, "stealthy"). The Requiem framework uses a combination of multiple deep learning models (that we refer to as "surrogates" and "spoofers") coupled with extensive, realistic simulations on a software-in-the-loop quadrotor UAV system. Requiem makes no assumptions about either the (types of) sensors or the onboard state estimation algorithm(s) -- it works so long as the latter is "learnable". We demonstrate the effectiveness of our system using various attacks across multiple missions as well as multiple sets of statistical analyses. We show that Requiem successfully exploits the modeling errors (i.e., causes significant deviations from planned mission parameters) while remaining stealthy (no detection even after {tens of meters of deviations}) and are generalizable (Requiem has potential to work across different attacks and sensor types).