Graph Neural Networks (GNNs) are a powerful tool for machine learning on graphs. GNNs combine node feature information with the graph structure by recursively passing neural messages along edges of the input graph. However, incorporating both graph structure and feature information leads to complex models and explaining predictions made by GNNs remains unsolved. Here we propose GNNEXPLAINER, the first general, model-agnostic approach for providing interpretable explanations for predictions of any GNN-based model on any graph-based machine learning task. Given an instance, GNNEXPLAINER identifies a compact subgraph structure and a small subset of node features that have a crucial role in GNN’s prediction. Further, GNNEXPLAINER can generate consistent and concise explanations for an entire class of instances. We formulate GNNEXPLAINER as an optimization task that maximizes the mutual information between a GNN’s prediction and distribution of possible subgraph structures. Experiments on synthetic and real-world graphs show that our approach can identify important graph structures as well as node features, and outperforms alternative baseline approaches by up to 43.0% in explanation accuracy. GNNEXPLAINER provides a variety of benefits, from the ability to visualize semantically relevant structures to interpretability, to giving insights into errors of faulty GNNs.
more »
« less
Semantically Equivalent Adversarial Rules for Debugging NLP models
Complex machine learning models for NLP are often brittle, making different predictions for input instances that are extremely similar semantically. To automatically detect this behavior for individual instances, we present semantically equivalent adversaries (SEAs) – semantic-preserving perturbations that induce changes in the model’s predictions. We generalize these adversaries into semantically equivalent adversarial rules (SEARs) – simple, universal replacement rules that induce adversaries on many instances. We demonstrate the usefulness and flexibility of SEAs and SEARs by detecting bugs in black-box state-of-the-art models for three domains: machine comprehension, visual question-answering, and sentiment analysis. Via user studies, we demonstrate that we generate high-quality local adversaries for more instances than humans, and that SEARs induce four times as many mistakes as the bugs discovered by human experts. SEARs are also actionable: retraining models using data augmentation significantly reduces bugs, while maintaining accuracy.
more »
« less
- Award ID(s):
- 1756023
- PAR ID:
- 10100478
- Date Published:
- Journal Name:
- Annual Meeting of the Association for Computational Linguistics (ACL)
- Page Range / eLocation ID:
- 856 to 865
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Localizing bugs in distributed applications is complicated by the potential presence of server/middleware misconfigurations and intermittent network connectivity. In this paper, we present a novel approach to localizing bugs in distributed web applications, targeting the important domain of full-stack JavaScript applications. The debugged application is first automatically refactored to create its semantically equivalent centralized version by gluing together the application’s client and server parts, thus separating the programmer-written code from configuration/environmental issues as suspected bug causes. The centralized version is then debugged to fix various bugs. Finally, based on the bug fixing changes of the centralized version, a patch is automatically generated to fix the original application source files. We show how our approach can be used to catch bugs that include performance bottlenecks and memory leaks. These results indicate that our debugging approach can facilitate the challenges of localizing and fixing bugs in web applications.more » « less
-
Humans are the final decision makers in critical tasks that involve ethical and legal concerns, ranging from recidivism prediction, to medical diagnosis, to fighting against fake news. Although machine learning models can sometimes achieve impressive performance in these tasks, these tasks are not amenable to full automation. To realize the potential of machine learning for improving human decisions, it is important to understand how assistance from machine learning models affects human performance and human agency. In this paper, we use deception detection as a testbed and investigate how we can harness explanations and predictions of machine learning models to improve human performance while retaining human agency. We propose a spectrum between full human agency and full automation, and develop varying levels of machine assistance along the spectrum that gradually increase the influence of machine predictions. We find that without showing predicted labels, explanations alone slightly improve human performance in the end task. In comparison, human performance is greatly improved by showing predicted labels (>20% relative improvement) and can be further improved by explicitly suggesting strong machine performance. Interestingly, when predicted labels are shown, explanations of machine predictions induce a similar level of accuracy as an explicit statement of strong machine performance. Our results demonstrate a tradeoff between human performance and human agency and show that explanations of machine predictions can moderate this tradeoff.more » « less
-
Container isolation is implemented through OS-level virtualization, such as Linux namespaces. Unfortunately, these mechanisms are extremely challenging to implement correctly and, in practice, suffer from functional interference bugs, which compromise container security. In particular, functional interference bugs allow an attacker to extract information from another container running on the same machine or impact its integrity by modifying kernel resources that are incorrectly isolated. Despite their impact, functional interference bugs in OS-level virtualization have received limited attention in part due to the challenges in detecting them. Instead of causing memory errors or crashes, many functional interference bugs involve hard-to-catch logic errors that silently produce semantically incorrect results. This paper proposes KIT, a dynamic testing framework that discovers functional interference bugs in OS-level virtualization mechanisms, such as Linux namespaces. The key idea of KIT is to detect inter-container functional interference by comparing the system call traces of a container across two executions, where it runs with and without the preceding execution of another container. To achieve high efficiency and accuracy, KIT includes two critical components: an efficient algorithm to generate test cases that exercise inter-container data flows and a system call trace analysis framework that detects functional interference bugs and clusters bug reports. KIT discovered 9 functional interference bugs in Linux kernel 5.13, of which 6 have been confirmed. All bugs are caused by logic errors, showing that this approach is able to detect hard-to-catch semantic bugs.more » « less
-
Language-level guarantees---like module runtime isolation for WebAssembly (Wasm)---are only as strong as the compiler that produces a final, native-machine-specific executable. The process of lowering language-level constructions to ISA-specific instructions can introduce subtle bugs that violate security guarantees. In this paper, we present Crocus, a system for lightweight, modular verification of instruction-lowering rules within Cranelift, a production retargetable Wasm native code generator. We use Crocus to verify lowering rules that cover WebAssembly 1.0 support for integer operations in the ARM aarch64 backend. We show that Crocus can reproduce 3 known bugs (including a 9.9/10 severity CVE), identify 2 previously-unknown bugs and an underspecified compiler invariant, and help analyze the root causes of a new bug.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.18653/v1/P18-1079
