The team is back in Magdeburg after the European Police Congress and is continuing its work on the project. The congress brought positive feedback, many interesting discussions and showed once again the need and the innovation possibilities in civil security research.
The team would like to thank the VDI and the BMBF for this wonderful opportunity!
Findings and prototypes of the EVOK project will be presented at the European Police Congress 2021 The congress will take place from September 14 to 15, 2021 at the bcc Berlin Congress Center. The European Police Congress is organized by Behörden Spiegel and is aimed at decision-makers from the police, security authorities and industry. You will find us at the stand of the civil security demand
Further information can be found here.
Our article "Guided Pattern Mining for API Misuse Detection by Change-Based Code Analysis" by Sebastian Nielebock, Robert Heumüller, Kevin Michael Schott and Frank Ortmeier has been accepted for publication in the journal Fachjournal "Automated Software Engineering“ published by Springer-Verlag.
In this article, we address the problem of API abuse, which occurs when developers incorrectly use the application programming interface of a particular code library. Existing detection tools use pattern mining techniques, i.e. they collect a set of potentially correct API applications and derive common usage patterns as correct specifications. In this way, violations of these specifications can be identified as API abuse. In this work, we analyzed the effects of different prior filtering and search strategies to obtain source code for the pattern mining step and were able to show that certain strategies improve pattern mining and misuse detection results regardless of the mining and detection technique applied.
A preprint of the article can be found on arXiv.
Our paper "Exploit Those Code Reviews! Bigger Data for Deeper Learning" has been accepted for the demonstrations track at ESEC/FSE 2021!
The paper presents the ETCR infrastructure for mining Modern Code Review (MCR) datasets from GitHub pull requests together with the ETCR Elasticsearch dataset of more than 40k pull requests, more than 231k comments and more than 15k different file revisions.
A preprint of the paper and the ETCR Elasticsearch artifact will be made available shortly.
Since the EVOK project (real-time on-site reconnaissance and mission monitoring) won first place in the "Most innovative projects in the field of applied interactive technologies" category of the 2020 Hugo Junkers Prize, public interest has continued to grow.
On June 25, 2021, the EVOK team presented the latest state of research to the press in a field test. The project was also able to impress with its topicality, importance and innovation.
The Chair of Software Engineering (CSE) has successfully started the summer semester 2021 with digital lectures on software engineering, specification technology and programming paradigms as well as a large number of seminars and student projects.
Students and lecturers met routinely on the online teaching platforms and video conferencing tools in what is now the third (partially) digital semester to start teaching after the Easter break.
Of course, we are available for our students as usual by email, video conference and, depending on the pandemic situation, also to a limited extent in person.
The EVOK project of our chair was virtually awarded the Hugo Junkers Prize on March 29, 2021. EVOK was awarded first place in the category Most Innovative Projects in Applied Interactive Technologies. The competition was looking for innovative projects that use their findings and initiatives to realize new approaches, create new markets or change existing ones. EVOK was able to prevail against a total of almost 100 competitors.
Our paper "AndroidCompass: A Dataset of Android Compatibility Checks in Code Repositories" by Sebastian Nielebock et al. has been accepted for presentation at the 2021 IEEE/ACM 18th International Conference on Mining Software Repositories (MSR) as part of the Data Showcase Track in May 2021. This paper describes AndroidCompass - a dataset of historical changes to Android compatibility checks in open source projects. We have also published a preprint as well as the dataset and its replication package.
Our paper "Learning to Boost the Efficiency of Modern Code Review" by Robert Heumüller has been accepted for presentation at the PhD Symposium of the 43rd International Conference on Software Engineering.
In his PhD project, Mr. Heumüller explores how AI models can learn to support or partially replace human code reviewers by leveraging large software repositories and current deep learning algorithms.
The CSE is gearing up for 2021 with a new high-performance compute system! Equipped with several Nvidia-A100 Data Center GPUs, we will continue to participate in innovative research and development in the field of machine learning and scientific computing in the new year.
Our focus is on the application of learning-based methods for the analysis of large code bases for automatic program understanding and code reviews with the aim of improving software quality, as well as the conception of more reliable deep learning methods to make them usable for industrial use cases in small and medium-sized companies.