Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observations

Golatkar, Aditya; Achille, Alessandro; Soatto, Stefano

doi:10.1007/978-3-030-58526-6_23

Cited by 66 publications

(69 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Efficiency. Our method is fast and highly efficient in comparison to the existing approaches [19,20]. The Fisher Forgetting [19] and NTK based forgetting [20] approaches require Hessian approximation which is computationally very expensive.…”

Section: Discussionmentioning

confidence: 99%

“…Our method is fast and highly efficient in comparison to the existing approaches [19,20]. The Fisher Forgetting [19] and NTK based forgetting [20] approaches require Hessian approximation which is computationally very expensive. It took us more than 2 hours to run Fisher forgetting [19] for 1-class unlearning in ResNet18 on CIFAR-10.…”

Section: Discussionmentioning

confidence: 99%

“…They also give an upper-bound on the amount of remaining information in the network [2] after forgetting by exploiting the stability of SGD. [20] extend this work to update the final activations of the model. They present a neural tangent ker-nel (NTK) based approximation of the training process and use it to estimate the updated network weights after forgetting.…”

Section: Related Workmentioning

confidence: 96%

“…The unlearning (also called selective forgetting, data deletion, or scrubbing) solutions presented in the literature are focused on simple learning algorithms such a linear/logistic regression [31], random forests [6], and kmeans clustering [16]. Initial work on forgetting in convolutional networks is presented in [19,20]. However, these methods are shown to be effective only on small scale problems and are computationally expensive.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fast Yet Effective Machine Unlearning

Tarun¹,

Chundawat²,

Mandal³

et al. 2021

Preprint

View full text Add to dashboard Cite

Unlearning the data observed during the training of a machine learning (ML) model is an important task that can play a pivotal role in fortifying the privacy and security of ML-based applications. This paper raises the following questions: (i) can we unlearn a class/classes of data from a ML model without looking at the full training data even once? (ii) can we make the process of unlearning fast and scalable to large datasets, and generalize it to different deep networks? We introduce a novel machine unlearning framework with error-maximizing noise generation and impair-repair based weight manipulation that offers an efficient solution to the above questions. An error-maximizing noise matrix is learned for the class to be unlearned using the original model. The noise matrix is used to manipulate the model weights to unlearn the targeted class of data. We introduce impair and repair steps for a controlled manipulation of the network weights. In the impair step, the noise matrix along with a very high learning rate is used to induce sharp unlearning in the model. Thereafter, the repair step is used to regain the overall performance. With very few update steps, we show excellent unlearning while substantially retaining the overall model accuracy. Unlearning multiple classes requires a similar number of update steps as for the single class, making our approach scalable to large problems. Our method is quite efficient in comparison to the existing methods, works for multi-class unlearning, doesn't put any constraints on the original optimization mechanism or network design, and works well in both small and large-scale vision tasks. This work is an important step towards fast and easy implementation of unlearning in deep networks. We will make the source code publicly available.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 96%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fast Yet Effective Machine Unlearning

Tarun¹,

Chundawat²,

Mandal³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…They usually employ gradient based update strategies to quickly eliminate the influence of samples that are requested to be deleted [40]. For example, Guo et al [26], Golatkar et al [24], and Golatkar et al [25] proposed different Newton's methods to approximate retraining for convex models, e.g., linear regression, logistic regression, and the last fully connected layer of a neural network. An alternative is to eliminate the influence of the samples that need to be deleted to the learned model based on 1 It is worth noting that the purpose of unlearning is different from differential privacy (DP) methods [18,20] which aim to protect users' privacy information instead of deleting them.…”

Section: Machine Unlearningmentioning

confidence: 99%

Recommendation Unlearning

Chen¹,

Sun²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Recommender systems provide essential web services by learning users' personal preferences from collected data. However, in many cases, systems also need to forget some training data. From the perspective of privacy, users desire a tool to erase the impacts of their sensitive data from the trained models. From the perspective of utility, if a system's utility is damaged by some bad data, the system needs to forget such data to regain utility. While unlearning is very important, it has not been well-considered in existing recommender systems. Although there are some researches have studied the problem of machine unlearning, existing methods can not be directly applied to recommendation as they are unable to consider the collaborative information.In this paper, we propose RecEraser, a general and efficient machine unlearning framework tailored to recommendation tasks. The main idea of RecEraser is to divide the training set into multiple shards and train submodels with these shards. Specifically, to keep the collaborative information of the data, we first design three novel data partition algorithms to divide training data into balanced groups. We then further propose an adaptive aggregation method to improve the global model utility. Experimental results on three public benchmarks show that RecEraser can not only achieve efficient unlearning but also outperform the state-of-the-art unlearning methods in terms of model utility. The source code can be found at https://github.com/chenchongthu/Recommendation-Unlearning CCS CONCEPTS• Information systems → Recommender systems; • Security and privacy → Privacy protections.

show abstract

Active forgetting via influence estimation for neural networks

Meng

Yang

Liu

et al. 2022

Int J of Intelligent Sys

View full text Add to dashboard Cite

The rapidly exploding of user data, especially applications of neural networks, involves analyzing data collected from individuals, which brings convenience to life. Meanwhile, privacy leakage in the applications 1% and 3% accuracy drop with more than 80% forgetting rate on average for logistic regression models and convolutional neural networks. The accuracy drop is reduced by 2%-3% compared to most state-of-the-art methods.

show abstract

Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observations

Cited by 66 publications

References 13 publications

Fast Yet Effective Machine Unlearning

Fast Yet Effective Machine Unlearning

Recommendation Unlearning

Active forgetting via influence estimation for neural networks

Contact Info

Product

Resources

About