ABSTRACT Machine learning techniques are increasingly used for high‐stakes decision‐making, such as college admissions, loan attribution, or recidivism prediction. Thus, it is crucial to ensure that the models learnt can be audited or understood by human users, do not create or reproduce discrimination or bias and do not leak sensitive information regarding their training data. Indeed, interpretability, fairness, and privacy are key requirements for the development of responsible machine learning, and all three have been studied extensively during the last decade. However, they were mainly considered in isolation, while in practice they interplay with each other, either positively or negatively. In this survey paper, we review the literature on the interactions between these three desiderata. More precisely, for each pairwise interaction, we summarize the identified synergies and tensions. These findings highlight several fundamental theoretical and empirical conflicts, while also demonstrating that jointly considering these different requirements is challenging when one aims at preserving a high level of utility. To solve this issue, we also discuss possible conciliation mechanisms, showing that a careful design can enable to successfully handle these different concerns in practice.

Traffic optimization and smart buffering are fundamental to achieve both great application performance and resource efficiency in data centers with heterogeneous workloads, including incast and elephant traffics. However, general performance models providing insights on how various factors affect traffic performance metrics needed by these management functions are missing. For the special case of incast, the existing models are analytical ones, either tightly coupled with a particular protocol version or specific to certain empirical data. Motivated by this observation, this paper proposes an SDN-enabled machine-learning-based performance modeling approach in data center networks that leverages random forest predictions. Evaluations based on datasets constructed through intensive NS-3 simulations show that we can achieve accurate predictions of incast and elephant performance metrics based on various features. With this performance modeling capability, smart buffering schemes or traffic optimization algorithms could anticipate and efficiently optimize system parameters adjustment to achieve optimal performance continuously in data centers.

As post hoc explanation methods are increasingly being leveraged to explain complex models in high-stakes settings, it becomes critical to ensure that the quality of the resulting explanations is consistently high across all subgroups of a population. For instance, it should not be the case that explanations associated with instances belonging to, e.g., women, are less accurate than those associated with other genders. In this work, we initiate the study of identifying group-based disparities in explanation quality. To this end, we first outline several key properties that contribute to explanation quality-namely, fidelity (accuracy), stability, consistency, and sparsity-and discuss why and how disparities in these properties can be particularly problematic. We then propose an evaluation framework which can quantitatively measure disparities in the quality of explanations. Using this framework, we carry out an empirical analysis with three datasets, six post hoc explanation methods, and different model classes to understand if and when group-based disparities in explanation quality arise. Our results indicate that such disparities are more likely to occur when the models being explained are complex and non-linear. We also observe that certain post hoc explanation methods (e.g., Integrated Gradients, SHAP) are more likely to exhibit disparities. Our work sheds light on previously unexplored ways in which explanation methods may introduce unfairness in real world decision making.

The use of black-box models (e.g., deep neural networks) in high-stakes decision-making systems, whose internal logic is complex, raises the need for providing explanations about their decisions. Model explanation techniques mitigate this problem by generating an interpretable and high-fidelity surrogate model (e.g., a logistic regressor or decision tree) to explain the logic of black-box models. In this work, we investigate the issue of fairwashing, in which model explanation techniques are manipulated to rationalize decisions taken by an unfair black-box model using deceptive surrogate models. More precisely, we theoretically characterize and analyze fairwashing, proving that this phenomenon is difficult to avoid due to an irreducible factor---the unfairness of the black-box model. Based on the theory developed, we propose a novel technique, called FRAUD-Detect (FaiRness AUDit Detection), to detect fairwashed models by measuring a divergence over subpopulation-wise fidelity measures of the interpretable model. We empirically demonstrate that this divergence is significantly larger in purposefully fairwashed interpretable models than in honest ones. Furthermore, we show that our detector is robust to an informed adversary trying to bypass our detector. The code implementing FRAUD-Detect is available at https://github.com/cleverhans-lab/FRAUD-Detect.

FairCORELS is an open-source Python module for building fair rule lists. It is a multi-objective variant of CORELS, a branch-and-bound algorithm to learn certifiably optimal rule lists. FairCORELS supports six statistical fairness metrics, proposes several exploration parameters and leverages on the fairness constraints to prune the search space efficiently. It can easily generate sets of accuracy-fairness trade-offs. The models learnt are interpretable by design and a sparsity parameter can be used to control their length.

Fairwashing refers to the risk that an unfair black-box model can be explained by a fairer model through post-hoc explanation manipulation. In this paper, we investigate the capability of fairwashing attacks by analyzing their fidelity-unfairness trade-offs. In particular, we show that fairwashed explanation models can generalize beyond the suing group (i.e., data points that are being explained), meaning that a fairwashed explainer can be used to rationalize subsequent unfair decisions of a black-box model. We also demonstrate that fairwashing attacks can transfer across black-box models, meaning that other black-box models can perform fairwashing without explicitly using their predictions. This generalization and transferability of fairwashing attacks imply that their detection will be difficult in practice. Finally, we propose an approach to quantify the risk of fairwashing, which is based on the computation of the range of the unfairness of high-fidelity explainers.

Black-box explanation is the problem of explaining how a machine learning model – whose internal logic is hidden to the auditor and generally complex – produces its outcomes. Current approaches for solving this problem include model explanation, outcome explanation as well as model inspection. While these techniques can be beneficial by providing interpretability, they can be used in a negative manner to perform fairwashing, which we define as promoting the false perception that a machine learning model respects some ethical values. In particular, we demonstrate that it is possible to systematically rationalize decisions taken by an unfair black-box model using the model explanation as well as the outcome explanation approaches with a given fairness metric. Our solution, LaundryML, is based on a regularized rule list enumeration algorithm whose objective is to search for fair rule lists approximating an unfair black-box model. We empirically evaluate our rationalization technique on black-box models trained on real-world datasets and show that one can obtain rule lists with high fidelity to the black-box model while being considerably less unfair at the same time.

Ridesharing, in which drivers offer to share their rides, allows reduction of travel costs for both drivers and riders; such practice is increasingly popular. Modern ridesharing systems, enhanced with location-based features, have improved user experience by enabling drivers and riders to arrange a trip in near real time. However, the fine-grained nature of location data collected by the service providers and exchanged between users raises privacy issues that could disrupt the adoption of such systems. In this paper, we present SRide: a privacy-preserving protocol for ridesharing that addresses the matching problem for dynamic ridesharing systems. We design and implement a prototype of SRide that operates in four steps. First, it generalizes users spatiotemporal data of users. Next, it relies on a secure filtering protocol to compute feasible matches. Then, it uses an improved version of Priv-2SP-SP- a privacy-preserving protocol to compute meeting points for ridesharing- to compute a ridesharing score for each feasible pair. Finally, it computes the optimal assignment of drivers and riders based on their ridesharing scores. We conduct an experimental trace-driven evaluation of the proposed scheme to demonstrate its practical feasibility.