No abstract

No abstract

In this paper, we study counterfactual fairness in text classification, which asks the question: How would the prediction change if the sensitive attribute referenced in the example were different? Toxicity classifiers demonstrate a counterfactual fairness issue by predicting that "Some people are gay" is toxic while "Some people are straight" is nontoxic. We offer a metric, counterfactual token fairness (CTF), for measuring this particular form of fairness in text classifiers, and describe its relationship with group fairness. Further, we offer three approaches, blindness, counterfactual augmentation, and counterfactual logit pairing (CLP), for optimizing counterfactual token fairness during training, bridging the robustness and fairness literature. Empirically, we find that blindness and CLP address counterfactual token fairness. The methods do not harm classifier performance, and have varying tradeoffs with group fairness. These approaches, both for measurement and optimization, provide a new path forward for addressing fairness concerns in text classification.

Deep neural networks have achieved state of the art accuracy at classifying molecules with respect to whether they bind to specific protein targets. A key breakthrough would occur if these models could reveal the fragment pharmacophores that are causally involved in binding. Extracting chemical details of binding from the networks could potentially lead to scientific discoveries about the mechanisms of drug actions. But doing so requires shining light into the black box that is the trained neural network model, a task that has proved difficult across many domains. Here we show how the binding mechanism learned by deep neural network models can be interrogated, using a recently described attribution method. We first work with carefully constructed synthetic datasets, in which the 'fragment logic' of binding is fully known. We find that networks that achieve perfect accuracy on held out test datasets still learn spurious correlations due to biases in the datasets, and we are able to exploit this non-robustness to construct adversarial examples that fool the model. The dataset bias makes these models unreliable for accurately revealing information about the mechanisms of protein-ligand binding. In light of our findings, we prescribe a test that checks for dataset bias given a hypothesis. If the test fails, it indicates that either the model must be simplified or regularized and/or that the training dataset requires augmentation.

Abstract. We define a small-step operational semantics for the ECMAScript standard language corresponding to JavaScript, as a basis for analyzing security properties of web applications and mashups. The semantics is based on the language standard and a number of experiments with different implementations and browsers. Some basic properties of the semantics are proved, including a soundness theorem and a characterization of the reachable portion of the heap.

A growing number of current web sites combine active content (applications) from untrusted sources, as in so-called mashups. The objectcapability model provides an appealing approach for isolating untrusted content: if separate applications are provided disjoint capabilities, a sound objectcapability framework should prevent untrusted applications from interfering with each other, without preventing interaction with the user or the hosting page. In developing language-based foundations for isolation proofs based on object-capability concepts, we identify a more general notion of authority safety that also implies resource isolation. After proving that capability safety implies authority safety, we show the applicability of our framework for a specific class of mashups. In addition to proving that a JavaScript subset based on Google Caja is capability safe, we prove that a more expressive subset of JavaScript is authority safe, even though it is not based on the object-capability model.

No abstract

Abstract. We give a new practical algorithm to compute, in finite time, a fixpoint (and often the least fixpoint) of a system of equations in the abstract numerical domains of zones and templates used for static analysis of programs by abstract interpretation. This paper extends previous work on the non-relational domain of intervals to relational domains. The algorithm is based on policy iteration techniques-rather than Kleene iterations as used classically in static analysis-and generates from the system of equations a finite set of simpler systems that we call policies. This set of policies satisfies a selection property which ensures that the minimal fixpoint of the original system of equations is the minimum of the fixpoints of the policies. Computing a fixpoint of a policy is done by linear programming. It is shown, through experiments made on a prototype analyzer, compared in particular to analyzers such as LPInv or the Octagon Analyzer, to be in general more precise and faster than the usual Kleene iteration combined with widening and narrowing techniques.