An official website of the United States government
Machine learning (ML) deployment projects are used by practitioners to automatically deploy ML models. While ML deployment projects aid practitioners, security vulnerabilities in these projects can make ML deployment infrastructure susceptible to security attacks. A systematic characterization of vulnerabilities can aid in identifying activities to secure ML deployment projects used by practitioners. We conduct an empirical study with 149 vulnerabilities mined from 12 open source ML deployment projects to characterize vulnerabilities in ML deployment projects. From our empirical study, we (i) find 68 of the 149 vulnerabilities are critically or highly severe; (ii) derive 10 consequences of vulnerabilities, e.g., unauthorized access to trigger ML deployments; and (iii) observe established quality assurance activities, such as code review to be used in the ML deployment projects. We conclude our paper by providing a set of recommendations for practitioners and researchers. Dataset used for our paper is available online.
more »
« less
Towards Automation for MLOps: An Exploratory Study of Bot Usage in Deep Learning Libraries
Machine learning (ML) operations or MLOps advocates for integration of DevOps- related practices into the ML development and deployment process. Adoption of MLOps can be hampered due to a lack of knowledge related to how development tasks can be automated. A characterization of bot usage in ML projects can help practitioners on the types of tasks that can be automated with bots, and apply that knowledge into their ML development and deployment process. To that end, we conduct a preliminary empirical study with 135 issues reported mined from 3 libraries related to deep learning: Keras, PyTorch, and Tensorflow. From our empirical study we observe 9 categories of tasks that are automated with bots. We conclude our work-in-progress paper by providing a list of lessons that we learned from our empirical study.
more »
« less
- Award ID(s):
- 2026869
- PAR ID:
- 10350204
- Date Published:
- Journal Name:
- 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC), 2022
- Page Range / eLocation ID:
- 1093 to 1097
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Of the hundreds of billions of dollars spent on developer wages, up to 25% accounts for fixing bugs. Companies like Google save significant human effort and engineering costs with automatic bug detection tools, yet automatically fixing them is still a nascent endeavour. Very recent work (including our own) demonstrates the feasibility of automatic program repair in practice. As automated repair technology matures, it presents great appeal for integration into developer workflows. We believe software bots are a promising vehicle for realizing this integration, as they bridge the gap between human software development and automated processes. We envision repair bots orchestrating automated refactoring and bug fixing. To this end, we explore what building a repair bot entails. We draw on our understanding of patch generation, validation, and real world software development interactions to identify six principles that bear on engineering repair bots and discuss related design challenges for integrating human workflows. Ultimately, this work aims to foster critical focus and interest for making repair bots a reality.more » « less
-
Bots have become critical for managing online communities on platforms, especially to match the increasing technical sophistication of online harms. However, community leaders often adoptthird-party bots, creating room for misalignment in their assumptions, expectations, and understandings (i.e., their technological frames) about them. On platforms where sharing bots can be extremely valuable, how community leaders can revise their frames about bots to more effectively adopt them is unclear. In this work, we conducted a qualitative interview study with 16 community leaders on Discord examining how they adopt third-party bots. We found that participants addressed challenges stemming from uncertainties about a bot's security, reliability, and fit through emergent social ecosystems. Formal and informal opportunities to discuss bots with others across communities enabled participants to revise their technological frames over time, closing gaps in bot-specific skills and knowledge. This social process of learning shifted participants' perspectives of the labor of bot adoption into something that was satisfying and fun, underscoring the value of collaborative and communal approaches to adopting bots. Finally, by shaping participants' mental models of the nature, value, and use of bots, social ecosystems also raise some practical tensions in how they support user creativity and customization in third-party bot use. Together, the social nature of adopting third-party bots in our interviews offers insight into how we can better support the sharing of valuable user-facing tools across online communities.more » « less
-
Bots are playing an increasingly important role in the creation of knowledge in Wikipedia. In many cases, editors and bots form tightly knit teams. Humans develop bots, argue for their approval, and maintain them, performing tasks such as monitoring activity, merging similar bots, splitting complex bots, and turning off malfunctioning bots. Yet this is not the entire picture. Bots are designed to perform certain functions and can acquire new functionality over time. They play particular roles in the editing process. Understanding these roles is an important step towards understanding the ecosystem, and designing better bots and interfaces between bots and humans. This is important for understanding Wikipedia along with other kinds of work in which autonomous machines affect tasks performed by humans. In this study, we use unsupervised learning to build a nine category taxonomy of bots based on their functions in English Wikipedia. We then build a multi-class classifier to classify 1,601 bots based on labeled data. We discuss different bot activities, including their edit frequency, their working spaces, and their software evolution. We use a model to investigate how bots playing certain roles will have differential effects on human editors. In particular, we build on previous research on newcomers by studying the relationship between the roles bots play, the interactions they have with newcomers, and the ensuing survival rate of the newcomers.more » « less
-
In this paper, we interrogate whether data quality issues track demographic characteristics such as sex, race and age, and whether automated data cleaning — of the kind commonly used in production ML systems — impacts the fairness of predictions made by these systems. To the best of our knowledge, the impact of data cleaning on fairness in downstream tasks has not been investigated in the literature.We first analyze the tuples flagged by common error detection strategies in five research datasets. We find that, while specific data quality issues, such as higher rates of missing values, are associated with membership in historically disadvantaged groups, poor data quality does not generally track demographic group membership. As a follow-up, we conduct a large-scale empirical study on the impact of automated data cleaning on fairness, involving more than 26,000 model evaluations on five datasets. We observe that, while automated data cleaning has an insignificant impact on both accuracy and fairness in the majority of cases, it is more likely to worsen fairness than to improve it, especially when the cleaning techniques are not carefully chosen. This finding is both significant and worrying, given that it potentially implicates many production ML systems. We make our code and experimental results publicly available.The analysis we conducted in this paper is difficult, primarily because it requires that we think holistically about disparities in data quality, disparities in the effectiveness of data cleaning methods, and impacts of such disparities on ML model performance for different demographic groups. Such holistic analysis can and should be supported with the help of data engineering research. Towards this goal, we envision the development of fairness-aware data cleaning methods, and their integration into complex pipelines for ML-based decision making.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/COMPSAC54236.2022.00171
