Source code and data relevant for the paper 'Learning-Based Testing the Sliding Window Behavior of TCP Implementations'

The dataset contains source code and data relevant for the paper "Learning-Based Testing the Sliding Window Behavior of TCP Implementations".Paper url: https://link.springer.com/chapter/10.1007/978-3-319-67113-0_12In this work, we use automata learning for register automata to infer models of TCP client implementations of Linux and FreeBSD. We then manually check the models for conformance to the RFC specifications. For the case study, we extended the register automata learner RALib so as to support data relations exhibited by TCP stacks (inequality over sums with constants). Note thatthis extension is mainly experimental.The dataset comprises the software components of our experimental setup apart from the actual implementations.In more concrete terms, the dataset contains:- the learner implementation - an implementation of the Java built RALib learning tool with the extensions this case study required, see https://bitbucket.org/learnlib/ralib/src/3812a20ae83b06436ad5fd74f1f5a57dc57b77bd?at=succ-expr- results - learned register automata models for TCP clients- experimental data - folders containing data about the experiments which lead to the learned modelNot included here is the framework used to learn TCP. This is the same as for the paper "Combining Model Learning and Model Checking to Analyze TCP Implementations", theonly difference is that we use a different learner. The framework is available at:https://gitlab.science.ru.nl/pfiteraubrostean/tcp-learner/tree/cav-aecWhat can be re-used:- the learner setup or RALib (to learn systems similar to TCP)- the learned models (if one wants to do something with the learned models)Also, the dataset is suitable for one who wants to try out/ extend learning for register automata.Paper Abstract: We develop a learning-based testing framework for register automaton models that can express the windowing behavior of TCP, thereby presenting the first significant application of register automata learning to realistic software for a class of automata with Boolean-arithmetic constraints over data values. We have applied our framework to TCP implementations belonging to different operating systems and have found a violation of the TCP specification in Linux and Windows. The violation has been confirmed by Linux developers.