The decentralized nature of federated learning makes detecting and defending against adversarial attacks a challenging task. This paper focuses on backdoor attacks in the federated learning setting, where the goal of the adversary is to reduce the performance of the model on targeted tasks while maintaining good performance on the main task. Unlike existing works, we allow non-malicious clients to have correctly labeled samples from the targeted tasks. We conduct a comprehensive study of backdoor attacks and defenses for the EMNIST dataset, a real-life, user-partitioned, and non-iid dataset. We observe that in the absence of defenses, the performance of the attack largely depends on the fraction of adversaries present and the "complexity'' of the targeted task. Moreover, we show that norm clipping and "weak'' differential privacy mitigate the attacks without hurting the overall performance. We have implemented the attacks and defenses in TensorFlow Federated (TFF), a TensorFlow framework for federated learning. In open-sourcing our code, our goal is to encourage researchers to contribute new attacks and defenses and evaluate them on standard federated datasets.
To improve real-world applications of machine learning, experienced modelers develop intuition about their datasets, their models, and how the two interact. Manual inspection of raw data - of representative samples, of outliers, of misclassifications - is an essential tool in a) identifying and fixing problems in the data, b) generating new modeling hypotheses, and c) assigning or refining human-provided labels. However, manual data inspection is problematic for privacy sensitive datasets, such as those representing the behavior of real-world individuals. Furthermore, manual data inspection is impossible in the increasingly important setting of federated learning, where raw examples are stored at the edge and the modeler may only access aggregated outputs such as metrics or model parameters. This paper demonstrates that generative models - trained using federated methods and with formal differential privacy guarantees - can be used effectively to debug many commonly occurring data issues even when the data cannot be directly inspected. We explore these methods in applications to text with differentially private federated RNNs and to images using a novel algorithm for differentially private federated GANs.
Privacy concerns have led to the development of privacy-preserving approaches for learning models from sensitive data. Yet, in practice, even models learned with privacy guarantees can inadvertently memorize unique training examples or leak sensitive features. To identify such privacy violations, existing model auditing techniques use finite adversaries defined as machine learning models with (a) access to some finite side information (e.g., a small auditing dataset), and (b) finite capacity (e.g., a fixed neural network architecture). Our work investigates the requirements under which an unsuccessful attempt to identify privacy violations by a finite adversary implies that no stronger adversary can succeed at such a task. We do so via parameters that quantify the capabilities of the finite adversary, including the size of the neural network employed by such an adversary and the amount of side information it has access to as well as the regularity of the (perhaps privacy-guaranteeing) audited model.
Local differential privacy (LDP) is a strong notion of privacy for individual users that often comes at the expense of a significant drop in utility. The classical definition of LDP assumes that all elements in the data domain are equally sensitive. However, in many applications, some symbols are more sensitive than others. This work proposes a context-aware framework of local differential privacy that allows a privacy designer to incorporate the application's context into the privacy definition. For binary data domains, we provide a universally optimal privatization scheme and highlight its connections to Warner's randomized response (RR) and Mangat's improved response. Motivated by geolocation and web search applications, for $k$-ary data domains, we consider two special cases of context-aware LDP: block-structured LDP and high-low LDP. We study discrete distribution estimation and provide communication-efficient, sample-optimal schemes and information-theoretic lower bounds for both models. We show that using contextual information can require fewer samples than classical LDP to achieve the same accuracy.
We present Generative Adversarial rePresentations (GAP) as a data-driven framework for learning censored and/or fair representations. GAP leverages recent advancements in adversarial learning to allow a data holder to learn universal representations that decouple a set of sensitive attributes from the rest of the dataset. Under GAP, finding the optimal mechanism? {decorrelating encoder/decorrelator} is formulated as a constrained minimax game between a data encoder and an adversary. We show that for appropriately chosen adversarial loss functions, GAP provides {censoring} guarantees against strong information-theoretic adversaries and enforces demographic parity. We also evaluate the performance of GAP on multi-dimensional Gaussian mixture models and real datasets, and show how a designer can certify that representations learned under an adversary with a fixed architecture perform well against more complex adversaries.
Recently, a parametrized class of loss functions called $\alpha$-loss, $\alpha \in [1,\infty]$, has been introduced for classification. This family, which includes the log-loss and the 0-1 loss as special cases, comes with compelling properties including an equivalent margin-based form which is classification-calibrated for all $\alpha$. We introduce a generalization of this family to the entire range of $\alpha \in (0,\infty]$ and establish how the parameter $\alpha$ enables the practitioner to choose among a host of operating conditions that are important in modern machine learning tasks. We prove that smaller $\alpha$ values are more conducive to faster optimization; in fact, $\alpha$-loss is convex for $\alpha \le 1$ and quasi-convex for $\alpha >1$. Moreover, we establish bounds to quantify the degradation of the local-quasi-convexity of the optimization landscape as $\alpha$ increases; we show that this directly translates to a computational slow down. On the other hand, our theoretical results also suggest that larger $\alpha$ values lead to better generalization performance. This is a consequence of the ability of the $\alpha$-loss to limit the effect of less likely data as $\alpha$ increases from 1, thereby facilitating robustness to outliers and noise in the training data. We provide strong evidence supporting this assertion with several experiments on benchmark datasets that establish the efficacy of $\alpha$-loss for $\alpha > 1$ in robustness to errors in the training data. Of equal interest is the fact that, for $\alpha < 1$, our experiments show that the decreased robustness seems to counteract class imbalances in training data.
We present $\alpha$-loss, $\alpha \in [1,\infty]$, a tunable loss function for binary classification that bridges log-loss ($\alpha=1$) and $0$-$1$ loss ($\alpha = \infty$). We prove that $\alpha$-loss has an equivalent margin-based form and is classification-calibrated, two desirable properties for a good surrogate loss function for the ideal yet intractable $0$-$1$ loss. For logistic regression-based classification, we provide an upper bound on the difference between the empirical and expected risk at the empirical risk minimizers for $\alpha$-loss by exploiting its Lipschitzianity along with recent results on the landscape features of empirical risk functions. Finally, we show that $\alpha$-loss with $\alpha = 2$ performs better than log-loss on MNIST for logistic regression.
State-of-the-art machine learning algorithms can be fooled by carefully crafted adversarial examples. As such, adversarial examples present a concrete problem in AI safety. In this work we turn the tables and ask the following question: can we harness the power of adversarial examples to prevent malicious adversaries from learning identifying information from data while allowing non-malicious entities to benefit from the utility of the same data? For instance, can we use adversarial examples to anonymize biometric dataset of faces while retaining usefulness of this data for other purposes, such as emotion recognition? To address this question, we propose a simple yet effective method, called Siamese Generative Adversarial Privatizer (SGAP), that exploits the properties of a Siamese neural network to find discriminative features that convey identifying information. When coupled with a generative model, our approach is able to correctly locate and disguise identifying information, while minimally reducing the utility of the privatized dataset. Extensive evaluation on a biometric dataset of fingerprints and cartoon faces confirms usefulness of our simple yet effective method.