Artificial Intelligence and machine learning (AI/ML) failures can be challenging to analyze. Detailed understanding of how these systems fail is a valuable first step in designing better AI/ML safety and security practices.

Almost 50 years ago, software reliability was defined with a hardware mindset. While the research community grew, its industry acceptance was muted. Rethinking the definitions of failures and faults will usher meaningful research and create value to the ...

We describe machine-learned classifiers that act as correctness properties for runtime verification of autonomous systems. Additionally, we show an advance collision detection system using such machine learning correctness properties that provides an ...

The 1980s and 1990s were a “golden age” for deep thinkers on the topic of software reliability models and testing. But today that thought leadership has disappeared. In this virtual roundtable we ask, “What happened?”

Virtualization, which allows the sharing and partitioning of underlying hardware, enables application scalability and consolidation. Its broad use brings relevant opportunities along with challenges that are crucial for the success of modern computer ...

Recent algorithms show how the availability of structural knowledge, such as symmetries, can significantly improve autonomous system verification in terms of both running time and sample complexity.

With the increasing ubiquity of the Internet of Things (IoT), some issues with satisfying quality requirements are emerging. This article discusses the current landscape and concerns about IoT quality requirements.

We humorously discuss 12 fallacies when focusing on compute performance that we have frequently observed in practice. We follow each with a recommendation to mitigate the danger and hope to contribute to good benchmarking etiquette for data science.

We quantify and derive a general model for the collaboration stability of human mobility and demonstrate its importance for networking applications. Our results demonstrate that collaboration opportunities are highly dependent on the context where they ...

We discuss the differences and commonalities between network technologies used in supercomputers and data centers and outline a path to convergence at multiple layers. We predict that emerging smart networking solutions will accelerate that convergence.

Measuring cyber resilience is vital to our security. In this article, we examine the importance of measuring cyber resilience and potential approaches to improve its effectiveness.

Artificial intelligence (AI) chips draw much attention for cyberphysical systems since AI chips are promising to realize edge AI computing. We introduce the chip architecture that enables energy-efficient computing and design tools for AI chips.

Lines of code (LOC) have been among the most popular measures for predicting software defects, but so far, their ultimate use has not been adequately understood. This article illustrates that LOC-based defect predictions are misleading, irrespective of ...

Tailoring cloud support for each autonomous-driving application would require maintaining multiple infrastructures, potentially resulting in low resource utilization, low performance, and high management overhead. To address this problem, the authors ...

The authors of a Computer article from 2002 reflect on their proposal to use networks on chips to address scalable communications on silicon VLSI chips.

Knowingly launching a flawed IT system that could harm society’s most vulnerable as collateral damage isn’t just callous—it’s administratively evil. Any IT professional should be ashamed to be part of such a project.

This article studies the interplay between the performance, energy, and reliability (PER) of parallel-computing systems. It describes methods supporting the meaningful cross-platform analysis of this interplay. These methods lead to the PER software tool, ...

This article describes the lessons learned, challenges faced, and innovations made in designing and implementing a high-availability private cloud for research computing.

Although it offers effective fault tolerance, three-way replication triples a system's purchase price. The natural redundancy of functionality across devices within the home, as well as usage scenarios, should be exploited in a new approach to IoT fault ...