Program Behavior Research Articles

DAEGEN: A Modular Compiler for Exploring Decoupled Spatial Accelerators

Specialized hardware accelerators, particularly those that are programmable and flexible to target multiple problems in their domain, have proven to provide orders of magnitude speedup and energy efficiency. However, their design requires extensive manual effort, due to the need for hardware-software codesign to balance the degree and forms of specialization to the domains or program behaviors of interest. This article provides the first steps towards one approach for automating much of these processes. The insight behind our work is to recognize that decoupled spatial architectures both define a rich design space with many tradeoffs for different kinds of applications, and also can be composed out of a simple set of well-defined primitives. Therefore, we propose a modular accelerator design framework, DAEGEN, a.k.a. Decoupled Access Excution Accelerator Generator. This article defines an initial compiler and architecture primitives, and we discuss key challenges.

Measuring health professions students’ teamwork behavior using peer assessment: Validation of an online tool

Abstract Teamwork, an interprofessional competency, lends itself toward behaviorally based assessments. Peer assessment within interprofessional education is one method to determine how an individual's team skills are perceived by others on a team. Peer assessment has been used in numerous learning contexts to evaluate professionalism, including the health sciences. Yet there appears to be few, if any reports, regarding peer assessment of learners collaborating as an interprofessional team. We used the Comprehensive Assessment of Team-Member Effectiveness (CATME), an online teamwork survey application to facilitate self and peer teamwork assessments within the context of a large, required, interprofessional service-learning program at the University of Florida. This paper describes the validation of CATME for use among health sciences students. Data from 1351 students was collected across two academic years, 2014–2015 and 2015–2016, related to three behavioral characteristics, and measures of team satisfaction and psychological safety. Statistical procedures employed included an exploratory factor analysis (EFA) and item analysis. Results provided evidence that 90% of the variance was explained by a single factor solution. Our validation provided evidence that CATME is a suitable tool for the formative evaluation of teamwork behaviors in clinical health science programs but could be further refined to improve itself usability.

Stepping OCaml

Steppers, which display all the reduction steps of a given program, are a novice-friendly tool for understanding program behavior. Unfortunately, steppers are not as popular as they ought to be; indeed, the tool is only available in the pedagogical languages of the DrRacket programming environment. We present a stepper for a practical fragment of OCaml. Similarly to the DrRacket stepper, we keep track of evaluation contexts in order to reconstruct the whole program at each reduction step. The difference is that we support effectful constructs, such as exception handling and printing primitives, allowing the stepper to assist a wider range of users. In this paper, we describe the implementation of the stepper, share the feedback from our students, and show an attempt at assessing the educational impact of our stepper.

Seclusion and Restraint of Students With Disabilities: A 1-Year Legal Review

News of teenager’s death in California shocked the special education community in early December 2018. Despite the Individuals With Disabilities Education Act’s provision for positive behavior supports for students with disabilities, these inidividuals continue to be secluded and restrained at disproportionately higher rates than their nondisabled peers. Using the Special Ed Connection® database, the authors identified 11 court cases from 2018 and analyzed their contents based on individualized education program (IEP) behavior components, type of restraints used, and prevalence of seclusion. Within these cases, all 12 included students were boys, four IEPs included behavior components, nine students were physically or mechanically restrained, and two were secluded. A review of the four cases that included the use of positive behavior supports is included.

Well-definedness and observational equivalence for inductive–coinductive programs

Abstract We define notions of well-definedness and observational equivalence for programs of mixed inductive and coinductive types. These notions are defined by means of tests formulas which combine structural congruence for inductive types and modal logic for coinductive types. Tests also correspond to certain evaluation contexts. We define a program to be well-defined if it is strongly normalizing under all tests, and two programs are observationally equivalent if they satisfy the same tests. We show that observational equivalence is sufficiently coarse to ensure that least and greatest fixed point types are initial algebras and final coalgebras, respectively. This yields inductive and coinductive proof principles for reasoning about program behaviour. On the other hand, we argue that observational equivalence does not identify too many terms, by showing that tests induce a topology that, on streams, coincides with usual topology induced by the prefix metric. As one would expect, observational equivalence is, in general, undecidable, but in order to develop some practically useful heuristics we provide coinductive techniques for establishing observational normalization and observational equivalence, along with up-to techniques for enhancing these methods.

Analyzing and Disentangling Interleaved Interrupt-Driven IoT Programs

In the Internet of Things (IoT) community, wireless sensor network (WSN) is a key technique to enable ubiquitous sensing of environments and provide reliable services to applications. WSN programs, typically interrupt-driven, implement the functionalities via the collaboration of interrupt procedure instances (IPIs, namely executions of interrupt processing logic). However, due to the complicated concurrency model of WSN programs, the IPIs are interleaved intricately and the program behaviors are hard to predicate from the source codes. Thus, to improve the software quality of WSN programs, it is significant to disentangle the interleaved executions and develop various IPI-based program analysis techniques, including offline and online ones. As the common foundation of those techniques, a generic efficient and real-time algorithm to identify IPIs is urgently desired. However, the existing instance-identification approach cannot satisfy the desires. In this paper, we first formally define the concept of IPI. Next, we propose a generic IPI-identification algorithm, and prove its correctness, real-time, and efficiency. We also conduct comparison experiments to illustrate that our algorithm is more efficient than the existing one in terms of both time and space. As the theoretical analyses and empirical studies exhibit, our algorithm provides the groundwork for IPI-based analyses of WSN programs in IoT environment.

DroidCat: Effective Android Malware Detection and Categorization via App-Level Profiling

Most existing Android malware detection and categorization techniques are static approaches, which suffer from evasion attacks, such as obfuscation. By analyzing program behaviors, dynamic approaches are potentially more resilient against these attacks. Yet existing dynamic approaches mostly rely on characterizing system calls which are subject to system-call obfuscation. This paper presents DroidCat, a novel dynamic app classification technique, to complement existing approaches. By using a diverse set of dynamic features based on method calls and inter-component communication (ICC) Intents without involving permission, app resources, or system calls while fully handling reflection, DroidCat achieves superior robustness than static approaches as well as dynamic approaches relying on system calls. The features were distilled from a behavioral characterization study of benign versus malicious apps. Through three complementary evaluation studies with 34 343 apps from various sources and spanning the past nine years, we demonstrated the stability of DroidCat in achieving high classification performance and superior accuracy compared with the two state-of-the-art peer techniques that represent both static and dynamic approaches. Overall, DroidCat achieved 97% F1-measure accuracy consistently for classifying apps evolving over the nine years, detecting or categorizing malware, 16%–27% higher than any of the two baselines compared. Furthermore, our experiments with obfuscated benchmarks confirmed higher robustness of DroidCat over these baseline techniques. We also investigated the effects of various design decisions on DroidCat’s effectiveness and the most important features for our dynamic classification. We found that features capturing app execution structure such as the distribution of method calls over user code and libraries are much more important than typical security features such as sensitive flows.

Fully Automated Differential Fault Analysis on Software Implementations of Block Ciphers

Differential Fault Analysis (DFA) is considered as the most popular fault analysis method. While there are techniques that provide a fault analysis automation on the cipher level to some degree, it can be shown that when it comes to software implementations, there are new vulnerabilities, which cannot be found by observing the cipher design specification.This work bridges the gap by providing a fully automated way to carry out DFA on assembly implementations of symmetric block ciphers. We use a customized data flow graph to represent the program and develop a novel fault analysis methodology to capture the program behavior under faults. We establish an effective description of DFA as constraints that are passed to an SMT solver. We create a tool that takes assembly code as input, analyzes the dependencies among instructions, automatically attacks vulnerable instructions using SMT solver and outputs the attack details that recover the last round key (and possibly the earlier keys). We support our design with evaluations on lightweight ciphers SIMON, SPECK, and PRIDE, and a current NIST standard, AES. By automated assembly analysis, we were able to find new efficient DFA attacks on SPECK and PRIDE, exploiting implementation specific vulnerabilities, and previously published DFA on SIMON and AES. Moreover, we present a novel DFA on multiplication operation that has never been shown for symmetric block ciphers before. Our experimental evaluation also shows reasonable execution times that are scalable to current cipher designs and can easily outclass the manual analysis. Moreover, we present a method to check the countermeasure-protected implementations in a way that helps implementers to decide how many rounds should be protected. We note that this is the first work that automatically carries out DFA on cipher implementations without any plaintext or ciphertext information and therefore, can be generally applied to any input data to the cipher.

MiC

Graphics processing units (GPUs) 1 have enjoyed increasing popularity in recent years, which benefits from, for example, general-purpose GPU (GPGPU) for parallel programs and new computing paradigms, such as the Internet of Things (IoT). GPUs hold great potential in providing effective solutions for big data analytics while the demands for processing large quantities of data in real time are also increasing. However, the pervasive presence of GPUs on mobile devices presents great challenges for GPGPU, mainly because GPGPU integrates a large amount of processor arrays and concurrent executing threads (up to hundreds of thousands). In particular, the root causes of performance loss in a GPGPU program can not be revealed in detail by current approaches. In this article, we propose MiC (Multi-level Characterization), a framework that comprehensively characterizes GPGPU kernels at the instruction, Basic Block (BBL), and thread levels. Specifically, we devise Instruction Vectors (IV) and Basic Blocks Vectors (BBV), a Thread Similarity Matrix (TSM), and a Divergence Flow Statistics Graph (DFSG) to profile information in each level. We use MiC to provide insights into GPGPU kernels through the characterizations of 34 kernels from popular GPGPU benchmark suites such as Compute Unified Device Architecture (CUDA) Software Development Kit (SDK), Rodinia, and Parboil. In comparison with Central Processing Unit (CPU) workloads, we conclude the key findings as follows: (1) There are comparable Instruction-Level Parallelism (ILP); (2) The BBL count is significantly smaller than CPU workloads—only 22.8 on average; (3) The dynamic instruction count per thread varies from dozens to tens of thousands and it is extremely small compared to CPU benchmarks; (4) The Pareto principle (also called 90/10 rule) does not apply to GPGPU kernels while it pervasively exists in CPU programs; (5) The loop patterns are dramatically different from those in CPU workloads; (6) The branch ratio is lower than that of CPU programs but higher than pure GPU workloads. In addition, we have also shown how TSM and DFSG are used to characterize the branch divergence in a visual way, to enable the analysis of thread behavior in GPGPU programs. In addition, we show an optimization case for a GPGPU kernel from the bottleneck identified through its characterization result, which improves 16.8% performance.

Computational inference of the transcriptional regulatory network of Candida glabrata.

Candida glabrata is a major cause of candidiasis and the second most frequent opportunistic yeast pathogen. Its infectious and antifungal mechanisms are globally regulated by the transcription systems of pathogenic fungi. In this study, we reconstructed the genome-scale transcriptional regulatory network (TRN) of C. glabrata, consisting of 6634 interactive relationships between 145 transcription factors and 3230 target genes, based on genomic and transcriptomic data. The C. glabrata TRN was found to have a typical topological structure and significant network cohesiveness. Moreover, this network could be functionally divided into several sub-networks, including networks involving carbon, nitrogen, growth-associated metabolic profiles, stress response to acidity, hyperosmosis, peroxidation, hypoxia and virulence. Furthermore, by integrating the genome-scale metabolic model of C. glabrata, six essential metabolites and eight related enzymes were systematically selected as drug targets. Overall, elucidation of the genome-scale TRN of C. glabrata has expanded our knowledge of the contents and structures of microbial regulatory networks and improved our understanding of the regulatory behaviors of growth, metabolism and gene expression programs in response to environmental stimuli.

Static correction of Maude programs with assertions

In this paper, we present a novel transformation method for Maude programs featuring both automatic program diagnosis and correction. The input of our method is a reference specification A of the program behavior that is given in the form of assertions together with an overly general program R whose execution might violate the assertions. Our correction technique translates R into a refined program R′ in which every computation is also a computation in R that satisfies the assertions of A. The technique is first formalized for topmost rewrite theories, and then we generalize it to larger classes of rewrite theories that support nested structured configurations. Our technique copes with infinite space states and does not require the knowledge of any failing run. We report experiments that assess the effectiveness of assertion-driven correction.

EdSketch: execution-driven sketching for Java

Sketching is a synthesis approach that allows users to provide high-level insights into a synthesis problem and let synthesis tools complete low-level details. Users write sketches—partial programs that have “holes” and provide test assertions as the correctness criteria. The sketching techniques fill the holes with code fragments such that the complete program satisfies all test assertions. Traditional techniques translate the sketching problem to propositional satisfiability formulas and leverage SAT solvers to generate programs with the desired functionality. While effective for a range of small well-defined domains, such translation-based approaches have a key limitation when applying to real applications: They require either translating all relevant libraries that are invoked directly or indirectly by the given sketch or creating models of those libraries, which requires much manual effort. This paper introduces execution-driven sketching, a novel approach for synthesizing Java programs with on-demand candidate generation. The key novelty of our work is to leverage runtime behavior to prune a large amount of search space. EdSketch explores the actual program behaviors in the presence of libraries and sketches small parts of real-world applications, which may use complex constructs of modern languages, such as reflection, native calls and File I/O. We further leverage a set of pruning strategies based on Java syntax to expedite the synthesis process. EdSketch embodies our approach in two forms: a stateful search based on the Java PathFinder model checker; and a stateless search based on re-execution inspired by the VeriSoft model checker. Experimental results show that EdSketch can complete some sketches that contain complex constructs in the presence of libraries, recursive procedures and advanced features like reflection. Without translating to SAT, EdSketch ’s performance compares well with the SAT-based Sketch system for a range of small but complex data structure subjects.

Developing Action-Taking Programs in Sustainable Consumption Education: Applying the Transtheoretical Model of Behavior Change (TTM)

How to motivate individual consumers to engage in consumption activities that contribute to sustainable development through action-taking education programs? This is the question this conceptual paper attempts to address. This paper proposes a definition of sustainable consumer behavior and applies a theory-based, conceptual framework to develop action-taking programs in sustainable consumption education. Sustainable consumer behavior is defined as consumer acts that consider current and future interests of themselves and others when they purchase and use market goods. The behavior change theory used here refers to the transtheoretical model of behavior change (TTM), a commonly used behavior change theory in health and finance. The paper demonstrates how to facilitate the development of sustainable consumer behavior in education programs based on principles outlined in TTM with specific examples. Implications for education policies and future research are also discussed.

FastSyzkaller: Improving Fuzz Efficiency for Linux Kernel Fuzzing

Linux is one of the most popular operating systems and enjoys a large number of subscribers on a world-wide scale. To ensure its stability and reliability, fuzzing is an effective method to trigger crashes hidden in the Linux kernel. In this paper, we apply N-Gram model to extract vulnerable program behaviours and dig out vulnerable patterns to guide the test case generation phase of the traditional fuzzing technique so as to improve the fuzzing efficiency from the Linux system call aspect. The experiment is implemented on the basis of a Google well-known open-source project Syzkaller, a fuzzer that targeted to the Linux kernel and other OS kernels as well. It is an unsupervised coverage-guided kernel fuzzer, set up with a manager, a fuzzer and an executor. This paper proposes FastSyzkaller which combines Syzkaller with N-Gram model in its fuzzer to optimize the test case generation process to improve the fuzzing efficiency. In 4 weeks, FastSyzkaller exposes 29 different crash types. However, Syzkaller only exposes 12 types. FastSyzkaller also produces crashes 3x faster than Syzkaller, also with 3x more unique crashes from the quantity perspective.

Buying free rewards: the impact of a points-plus-cash promotion on purchase and reward redemption

While many studies have investigated consumer purchase behavior in reward programs, a better understanding of customer redemption behavior is lacking, particularly when promotions affect a core aspect of reward programs—free rewards. In this paper, we examine the impact of a promotion on purchase and reward redemption in a reward program in which consumers can partially cover the cost of a free reward with their money. The literature on reward programs suggests a positive reinforcement caused by reward redemption, whereas the literature on promotion provides different views regarding the existence of a postpromotion dip. Using data from a major retailer’s reward program, we verify that such a promotion attracted customers with less transaction activity and shorter tenure. Interestingly, consumers using the promotion increased their preference for hedonic rewards compared to their previously observed behavior. This change in preference persisted after the promotion ended. Overall, the promotion significantly increased the number of redemptions but generated a negative impact on subsequent consumer behavior by decreasing purchase incidence and quantity. Our findings point to a need to understand the trade-off between spending money on buying an otherwise free reward and future regular purchases.

Buying Free Rewards: The Impact of a Points-Plus-Cash Promotion on Purchase and Reward Redemption

While many studies have investigated consumer purchase behavior in reward programs, a better understanding of customer redemption behavior is lacking, particularly when promotions affect a core aspect of reward programs—free rewards. In this paper, we examine the impact of a promotion on purchase and reward redemption in a reward program in which consumers can partially cover the cost of a free reward with their money. The literature on reward programs suggests a positive reinforcement caused by reward redemption, whereas the literature on promotion provides different views regarding the existence of a postpromotion dip. Using data from a major retailer’s reward program, we verify that such a promotion attracted customers with less transaction activity and shorter tenure. Interestingly, consumers using the promotion increased their preference for hedonic rewards compared to their previously observed behavior. This change in preference persisted after the promotion ended. Overall, the promotion significantly increased the number of redemptions but generated a negative impact on subsequent consumer behavior by decreasing purchase incidence and quantity. Our findings point to a need to understand the trade-off between spending money on buying an otherwise free reward and future regular purchases.

Babylonian-style Programming: Design and Implementation of an Integration of Live Examples into General-purpose Source Code

When working on a program, developers traditionally have to simulate the behavior of the abstract code in their heads until they can execute the application. Live programming aims to support the development and comprehension of programs by providing more immediate feedback on program behavior, but the divide between code and behavior often remains. The goal of example-based live programming is to remove this gap by allowing programmers to explore the actual behavior of their code during development. This is achieved by defining live examples for parts of the application. The idea of live examples has been already addressed in other tools and environments. However, most of those solutions are limited to specific domains and are suitable only for small programs. Thus, we aim to extend the application of example-based live programming to more complex programs potentially spanning multiple modules. We investigate existing solutions to derive a set of requirements for an integration of live examples into source code. Based on these requirements we propose a new approach to live examples and present a prototype in its support. We reproduce, discuss, and extend scenarios from related work to show the practicality of our approach in the context of larger, more complicated, and with that also more realistic scenarios. Also, we measure and evaluate the system response time of our prototypical implementation. Our first results show that example-based live programming can provide more insights into the run-time behavior of parameterized code for non-trivial programs. They also reveal unsolved and new challenges affecting example-based live programming environments. In presenting this more general approach to example-based live programming, we hope to motivate further research into this area and to make practical solutions available.

Event Loops as First-Class Values: A Case Study in Pedagogic Language Design

The World model is an existing functional input-output mechanism for event-driven programming. It is used in numerous popular textbooks and curricular settings. The World model conflates two different tasks -- the definition of an event processor and its execution -- into one. This conflation imposes a significant (even unacceptable) burden on student users in several educational settings where we have tried to use it, e.g., for teaching physics. While it was tempting to pile on features to address these issues, we instead used the Scheme language design dictum of removing weaknesses that made them seem necessary. By separating the two tasks above, we arrived at a slightly different primitive, the reactor, as our basis. This only defines the event processor, and a variety of execution operators dictate how it runs. The new design enables programmatic control over event-driven programs. This simplifies reflecting on program behavior, and eliminates many unnecessary curricular dependencies imposed by the old design. This work has been implemented in the Pyret programming language. The separation of concerns has enabled new curricula, such as the Bootstrap:Physics curriculum, to take flight. Thousands of students use this new mechanism every year. We believe that reducing impedance mismatches improves their educational experience.

May the Interview Be With You: Signal Your Preferences.

May the Interview Be With You: Signal Your Preferences.

Type qualifiers as composable language extensions for code analysis and generation

The work presented here reformulates type qualifiers as composable language extensions that can be automatically and reliably composed by the end-user programmer. Type expressions can be annotated with type qualifiers to specify new subtyping relations that are expressive enough to detect many kinds of programming errors. Type qualifiers, as illustrated in our ableC extensible language framework for C, can also introduce rich forms of concrete syntax, can generate dynamic checks on data when static checks are infeasible or not appropriate, and can inject generated code that affects program behavior, for example to log or display program execution or for the run-time conversion of data.The ableC framework and extensions to it are implemented using context-free grammars and attribute grammars. This provides an expressive mechanism for type qualifier implementations to check for additional errors, e.g.dereferences to pointers not qualified by a “nonnull” qualifier, and generate custom and informative error messages. This approach distinguishes programmers that use language extensions from language engineers that develop extensions. The framework provides modular analyses that extension developers use to ensure that their extension will compose with other extensions that all pass these analyses. Thus, when a programmer selects a set of extensions to use they will automatically and reliably compose to form a working translator for the extended language.