Deep neural networks (DNNs) are vulnerable to backdoor attacks. Previous works have shown it extremely challenging to unlearn the undesired backdoor behavior from the network, since the entire network can be affected by the backdoor samples. In this paper, we propose a brand-new backdoor defense strategy, which makes it much easier to remove the harmful influence of backdoor samples from the model. Our defense strategy, Trap and Replace, consists of two stages. In the first stage, we bait and trap the backdoors in a small and easy-to-replace subnetwork. Specifically, we add an auxiliary image reconstruction head on top of the stem network shared with a light-weighted classification head. The intuition is that the auxiliary image reconstruction task encourages the stem network to keep sufficient low-level visual features that are hard to learn but semantically correct, instead of overfitting to the easy-to-learn but semantically incorrect backdoor correlations. As a result, when trained on backdoored datasets, the backdoors are easily baited towards the unprotected classification head, since it is much more vulnerable than the shared stem, leaving the stem network hardly poisoned. In the second stage, we replace the poisoned light-weighted classification head with an untainted one, by re-training it from scratchmore »
Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks
Deep neural networks (DNNs) provide excellent performance across a wide range of classification tasks, but their training requires high computational resources and is often outsourced to third parties. Recent work has shown that outsourced training introduces the risk that a malicious trainer will return a backdoored DNN that behaves normally on most inputs but causes targeted misclassifications or degrades the accuracy of the network when a trigger known only to the attacker is present. In this paper, we provide the first effective defenses against backdoor attacks on DNNs. We implement three backdoor attacks from prior work and use them to investigate two promising defenses, pruning and fine-tuning. We show that neither, by itself, is sufficient to defend against sophisticated attackers. We then evaluate fine-pruning, a combination of pruning and fine-tuning, and show that it successfully weakens or even eliminates the backdoors, i.e., in some cases reducing the attack success rate to 0% with only a 0.4% drop in accuracy for clean (non-triggering) inputs. Our work provides the first step toward defenses against backdoor attacks in deep neural networks.
- Award ID(s):
- 1801495
- Publication Date:
- NSF-PAR ID:
- 10084748
- Journal Name:
- Research in Attacks, Intrusions, and Defenses
- Page Range or eLocation-ID:
- 273 - 294
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The pervasiveness of neural networks (NNs) in critical computer vision and image processing applications makes them very attractive for adversarial manipulation. A large body of existing research thoroughly investigates two broad categories of attacks targeting the integrity of NN models. The first category of attacks, commonly called Adversarial Examples, perturbs the model's inference by carefully adding noise into input examples. In the second category of attacks, adversaries try to manipulate the model during the training process by implanting Trojan backdoors. Researchers show that such attacks pose severe threats to the growing applications of NNs and propose several defenses against each attack type individually. However, such one-sided defense approaches leave potentially unknown risks in real-world scenarios when an adversary can unify different attacks to create new and more lethal ones bypassing existing defenses. In this work, we show how to jointly exploit adversarial perturbation and model poisoning vulnerabilities to practically launch a new stealthy attack, dubbed AdvTrojan. AdvTrojan is stealthy because it can be activated only when: 1) a carefully crafted adversarial perturbation is injected into the input examples during inference, and 2) a Trojan backdoor is implanted during the training process of the model. We leverage adversarial noise in themore »
-
Federated learning (FL) is known to be susceptible to model poisoning attacks in which malicious clients hamper the accuracy of the global model by sending manipulated model updates to the central server during the FL training process. Existing defenses mainly focus on Byzantine-robust FL aggregations, and largely ignore the impact of the underlying deep neural network (DNN) that is used to FL training. Inspired by recent findings on critical learning periods (CLP) in DNNs, where small gradient errors have irrecoverable impact on the final model accuracy, we propose a new defense, called a CLP-aware defense against poisoning of FL (DeFL). The key idea of DeFL is to measure fine-grained differences between DNN model updates via an easy-to-compute federated gradient norm vector (FGNV) metric. Using FGNV, DeFL simultaneously detects malicious clients and identifies CLP, which in turn is leveraged to guide the adaptive removal of detected malicious clients from aggregation. As a result, DeFL not only mitigates model poisoning attacks on the global model but also is robust to detection errors. Our extensive experiments on three benchmark datasets demonstrate that DeFL produces significant performance gain over conventional defenses against state-of-the-art model poisoning attacks.
-
Several recent works have demonstrated highly effective model stealing (MS) attacks on Deep Neural Networks (DNNs) in black-box settings, even when the training data is unavailable. These attacks typically use some form of Out of Distribution (OOD) data to query the target model and use the predictions obtained to train a clone model. Such a clone model learns to approximate the decision boundary of the target model, achieving high accuracy on in-distribution examples. We propose Ensemble of Diverse Models (EDM) to defend against such MS attacks. EDM is made up of models that are trained to produce dissimilar predictions for OOD inputs. By using a different member of the ensemble to service different queries, our defense produces predictions that are highly discontinuous in the input space for the adversary's OOD queries. Such discontinuities cause the clone model trained on these predictions to have poor generalization on in-distribution examples. Our evaluations on several image classification tasks demonstrate that EDM defense can severely degrade the accuracy of clone models (up to 39.7%). Our defense has minimal impact on the target accuracy, negligible computational costs during inference, and is compatible with existing defenses for MS attacks.
-
Despite their tremendous success in a range of domains, deep learning systems are inherently susceptible to two types of manipulations: adversarial inputs -- maliciously crafted samples that deceive target deep neural network (DNN) models, and poisoned models -- adversely forged DNNs that misbehave on pre-defined inputs. While prior work has intensively studied the two attack vectors in parallel, there is still a lack of understanding about their fundamental connections: what are the dynamic interactions between the two attack vectors? what are the implications of such interactions for optimizing existing attacks? what are the potential countermeasures against the enhanced attacks? Answering these key questions is crucial for assessing and mitigating the holistic vulnerabilities of DNNs deployed in realistic settings. Here we take a solid step towards this goal by conducting the first systematic study of the two attack vectors within a unified framework. Specifically, (i) we develop a new attack model that jointly optimizes adversarial inputs and poisoned models; (ii) with both analytical and empirical evidence, we reveal that there exist intriguing "mutual reinforcement" effects between the two attack vectors -- leveraging one vector significantly amplifies the effectiveness of the other; (iii) we demonstrate that such effects enable a large designmore »
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Publisher's Version of Record
- https://doi.org/10.1007/978-3-030-00470-5_13
