Denial of Deadline: Network-Driven Accuracy Collapse in Distributed Inference Pipelines
2026-07-27 • Networking and Internet Architecture
Networking and Internet ArchitectureArtificial IntelligenceCryptography and SecurityDistributed, Parallel, and Cluster Computing
AI summaryⓘ
The authors explain that modern AI systems often use a fast method for quick answers and a slower, more accurate method that runs on powerful remote servers. They discovered a new weak spot where attackers can send bursty workloads to overwhelm the slow method, causing its results to be too late and ignored, which hurts overall accuracy. Through experiments in self-driving car tracking, they showed how these attacks can significantly reduce prediction quality, especially for rare objects like stop signs. Their work highlights the need to develop protections against such timing-based attacks in these combined inference systems.
inference systemsfast pathslow pathlatencyworkload attacksmulti-object trackingaccuracy collapseedge-cloud computingresource contentionrouting
Authors
Jhonatan Tavori, Gur-Eyal Sela, Ion Stoica, Gil Zussman
Abstract
Inference systems increasingly combine a fast path that returns predictions within the application's latency deadline together with a higher-accuracy slow path that runs higher-compute methods on stronger, remote hardware, so its results can be returned on time and combined with the fast path predictions. Across several application domains, we abstract this inference architecture as a fast path, a slow path, and a coordination layer with two functions: a router that invokes the slow path and a merger that decides whether to incorporate its returned predictions. In this work, we show that this new coordination layer exposes a new attack surface: shaped workload attacks, e.g., Yo-Yo bursts, can exploit contention at shared resources along the slow path to push benign users' slow-path predictions past their latency deadlines. The merger then discards those predictions, while the fast path continues to return timely outputs. We refer to the resulting loss of slow-path accuracy benefits as accuracy collapse. We demonstrate accuracy collapse in a two-tier edge-cloud multi-object tracking pipeline in autonomous driving. In simulation, approximately 4,000 burst-shaped requests increase benign p99 latency from 92ms to 2s, nearly eliminating the benefit of the slow path's cloud inference, reducing object tracking quality by 7.0 HOTA points on average. We further find that accuracy degradation can significantly vary (2.0-18.7 HOTA points), depending on the video intervals that are targeted in the attack, and that certain rare classes (e.g., stop signs) lose nearly half of their pre-attack prediction accuracy. These results show that workload attacks can degrade prediction quality without needing either access to model weights or victim data, and motivate research on attacks and defenses for routing, merging, scheduling, and resource isolation in these emerging inference pipeline architectures.