BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain

TLDR

Deep learning achieves state‑of‑the‑art performance across many tasks, yet its high training cost forces users to outsource training or rely on pre‑trained models. The study demonstrates that outsourcing training enables adversaries to embed backdoors—BadNets—that perform normally on standard data yet misclassify attacker‑chosen inputs. We construct a backdoored handwritten‑digit classifier and a U.S. street‑sign detector that mislabels stop signs as speed limits when a sticker is applied, showing the backdoor survives retraining and reduces accuracy by about 25 % when triggered.

Abstract

Deep learning-based techniques have achieved state-of-the-art performance on a wide variety of recognition and classification tasks. However, these networks are typically computationally expensive to train, requiring weeks of computation on many GPUs; as a result, many users outsource the training procedure to the cloud or rely on pre-trained models that are then fine-tuned for a specific task. In this paper we show that outsourced training introduces new security risks: an adversary can create a maliciously trained network (a backdoored neural network, or a \emph{BadNet}) that has state-of-the-art performance on the user's training and validation samples, but behaves badly on specific attacker-chosen inputs. We first explore the properties of BadNets in a toy example, by creating a backdoored handwritten digit classifier. Next, we demonstrate backdoors in a more realistic scenario by creating a U.S. street sign classifier that identifies stop signs as speed limits when a special sticker is added to the stop sign; we then show in addition that the backdoor in our US street sign detector can persist even if the network is later retrained for another task and cause a drop in accuracy of {25}\% on average when the backdoor trigger is present. These results demonstrate that backdoors in neural networks are both powerful and---because the behavior of neural networks is difficult to explicate---stealthy. This work provides motivation for further research into techniques for verifying and inspecting neural networks, just as we have developed tools for verifying and debugging software.