Knowledge-based relevance filtering for efficient system-level test-based model generation

Lecture Notes in Computer Science

Jönsson

Vaandrager

2019

Model learning is a black-box technique for constructing state machine models of software and hardware components, which has been successfully used in areas such as telecommunication, banking cards, network protocols, and control software. The underlying theoretic framework (active automata learning) was first introduced in a landmark paper by Dana Angluin in 1987 for finite state machines. In order to make model learning more widely applicable, it must be further developed to scale better to large models and to generate richer classes of models. Recently, various techniques have been employed to extend automata learning to extended automata models, which combine control flow with guards and assignments to data variables. Such techniques infer guards over data parameters and assignments from observations of test output. In the black-box model of active automata learning this can be costly and require many tests, while in many application scenarios source code is available for analysis. In this paper, we explore some directions for future research on how black-box model learning can be enhanced using white-box information extraction methods, with the aim to maintain the benefits of dynamic black-box methods while making effective use of information that can be obtained through white-box techniques.

Section: Related Work and Applicationsmentioning

confidence: 99%

Combining Black-Box and White-Box Techniques for Learning Register Automata

Lecture Notes in Computer Science

Jönsson

Vaandrager

2019

“…In [28,52] optimizations are discussed to classic learning algorithms that aim at saving membership queries in practical scenarios. Additionally, the use of filters (exploiting domain specific expert knowledge) has been proven as a practical solution to the problem [42]. C: Parameters and value domains Active learning classically is based on abstract communication alphabets.…”

Section: Challenges In Practical Applicationsmentioning

confidence: 99%

Introduction to Active Automata Learning from a Practical Perspective

Steffen

Formal Methods for Eternal Networked Software Systems

Merten

2011

Self Cite

123

Abstract. In this chapter we give an introduction to active learning of Mealy machines, an automata model particularly suited for modeling the behavior of realistic reactive systems. Active learning is characterized by its alternation of an exploration phase and a testing phase. During exploration phases so-called membership queries are used to construct hypothesis models of a system under learning. In testing phases so-called equivalence queries are used to compare respective hypothesis models to the actual system. These two phases are iterated until a valid model of the target system is produced. We will step-wisely elaborate on this simple algorithmic pattern, its underlying correctness arguments, its limitations, and, in particular, ways to overcome apparent hurdles for practical application. This should provide students and outsiders of the field with an intuitive account of the high potential of this challenging research area in particular concerning the control and validation of evolving reactive systems. MotivationInteroperability remains a fundamental challenge when connecting heterogeneous systems [10]. The Connect Integrated Project [35] aims at overcoming the interoperability barrier by synthesizing required Connectors on the fly in five steps [5,21]: (i) extracting knowledge from, (ii) learning about and (iii) reasoning about the interaction behavior of networked systems, as a basis for (iv) synthesizing Connectors [33,34], and subsequently (v) generating and deploying them for immediate use [9].This chapter focuses on the foundations for step (ii), namely on techniques for leveraging and enriching the extracted knowledge by means of experimentation with the targeted components. Central are here advanced techniques for active automata learning [3,38,4,31,50], which are designed for optimally aggregating, and where necessary completing, the observed behavior.Characteristic for active learning automata learning is its iterative alternation between a "testing" phase for completing the transitions relation of the model aggregated from the observed behavior, and an equivalence checking phase, which This work is supported by the European FP 7 project CONNECT (IST 231167).

“…The additional "new" data value is used as data parameter. We have used symmetry reduction, i.e., normalizing the order of data values occurring in an input word as described in [13], to reduce the number of queries when inferring plain Mealy machine models.…”

Section: Experimental Evaluationmentioning

confidence: 99%

Inferring Semantic Interfaces of Data Structures

Leveraging Applications of Formal Methods, Verification and Validation. Technologies for Mastering Change

Isberner

Steffen

et al. 2012

Self Cite

Abstract. In this paper, we show how to fully automatically infer semantic interfaces of data structures on the basis of systematic testing. Our semantic interfaces are a generalized form of Register Automata (RA), comprising parameterized input and output, allowing to model control-and data-flow in component interfaces concisely. Algorithmic key to the automated synthesis of these semantic interfaces is the extension of an active learning algorithm for Register Automata to explicitly deal with output. We evaluated our algorithm on a complex data structure, a "stack of stacks", the largest of which we could learn in merely 20 seconds with less than 4000 membership queries, resulting in a model with rougly 800 nodes. In contrast, even when restricting the data domain to just four values, the corresponding plain Mealy machine would have more than 10 9 states and presumably require billions of membership queries.