Alin Stefanescu

University of Bucharest | Unibuc · Department of Computer Science

Natural language processing (NLP) applications are becoming increasingly popular today, largely due to recent advances in theory (machine learning and knowledge representation) and the computational power required to train and store large language models and data. Since NLP applications such as Alexa, Google Assistant, Cortana, Siri, and chatGPT ar...

The requirements of today’s applications and their users set high demands and expectations. AI is a part of these and has played an important role recently. However, the credibility of AI methods is controversial in many cases, as is data security and user privacy. On the other hand, Blockchain is a trending technology that offers security and priv...

The increasingly popular no- or low-code paradigm is based on functional blocks connected on a graphical interface that is accessible to many stakeholders in an application. Areas such as machine learning, DevOps, digital twins, simulations, and video games use this technique to facilitate communication between stakeholders regarding the business l...

Robotic Process Automation has recently been used in many fields to automate business-oriented processes. Industries such as finance, transportation, and retail report significant return on investment (ROI) after replacing redundant, repetitive, and error-prone work performed by human workers with RPA software agents. In our research, we found that...

Recent work has shown how easily white-box adversarial attacks can be applied to state-of-the-art image classifiers. However, real-life scenarios resemble more the black-box adversarial conditions, lacking transparency and usually imposing natural, hard constraints on the query budget. It is usual for the black-box adversarial literature to assume...

IoT systems are becoming more and more present in our lives and the demand recently increased with the explosion of external services offered by systems such as healthcare, smart city management or various providers. However, the connection of private IoT driven home systems and passing data to these external services can pose significant privacy i...

Traffic light control optimization is nowadays an important part of a smart city, given the advancement of sensors, IoT, and edge computing capabilities. The optimization method targeted by our work follows a general trend in the community: dynamically switching traffic light phases depending on the current traffic state. Reinforcement learning was...

As is the case with any very complex and interactive software, many video games are released with various minor or major issues that can potentially affect the user experience, cause security issues for players, or exploit the companies that deliver the products. To test their games, companies invest important resources in quality assurance personn...

In this paper,we propose GameBlockchain, an open-source blockchain framework designed to support secure transactions of NFTs in modern computer games. Its purpose is to enable game industry stakeholders such as game developers, content creators, and regular gamers to create and exchange game assets in a more secure and trusted environment. The secu...

We propose EvoBA (All the work is open source: https://github.com/andreiilie1/BBAttacks A full paper version is available at https://arxiv.org/abs/2107.05754), a black-box adversarial attack based on a surprisingly simple evolutionary search strategy. EvoBA is query-efficient, minimizes 𝐿0 adversarial perturbations, and does not require any form of...

Video game development is a growing industry nowadays with high revenues. However, even if there are many resources invested in the software development process, many games still contain bugs or performance issues that affect the user experience. This paper presents ideas on how computer vision methods can be used to automate the process of game te...

Recent work has shown how easily white-box adversarial attacks can be applied to state-of-the-art image classifiers. However, real-life scenarios resemble more the black-box adversarial conditions, lacking transparency and usually imposing natural, hard constraints on the query budget. We propose $\textbf{EvoBA}$, a black-box adversarial attack bas...

Combining fuzzing techniques and reinforcement learning could be an important direction in software testing. However, there is a gap in support for experimentation in this field, as there are no open-source tools to let academia and industry to perform experiments easily. The purpose of this paper is to fill this gap by introducing a new framework,...

Robotic Process Automation (RPA) is a technology that has grown tremendously in the last years, due to its usability in the area of process automation. An essential part of any software development process is quality assurance, so testing will be very important for RPA processes. However, the classical software techniques are not always suitable fo...

Recently, academics and professionals proposed a substantial number of identity management solutions based on blockchain technology. Among them, uPort appeared promising, being considered for both personal and administrative operations. This paper analyzes the open-source version uPort Open in terms of risk delegation and private key recovery of th...

Characterizing how effective a machine learning algorithm is while being trained and tested on slightly different data is a widespread matter. The property of models which perform well under this general framework is commonly known as robustness. We propose a class of model-agnostic empirical ro-bustness measures for image classification tasks. To...

In the last years, identity management solutions on blockchain were proposed as a possible solution to the digital identity management problem. However, they are still at an early stage and further research needs to be done to conclude whether identity systems could benefit from the use of blockchain or not. Motivated by this, we investigate identi...

In the last years, identity management solutions on blockchain were proposed as a possible solution to digital identity management problems. However, they are still at an early stage, and further research needs to be done to conclude whether identity systems could benefit from the use of blockchain or not. Motivated by this, we investigate blockcha...

This paper presents an improvement to a new open-source testing tool capable of performing concolic execution on x86 binaries. The novelty is to use a reinforcement learning solution that reduces the number of symbolically executed states. It does so by learning a set of models that predict how efficiently it would be to change the conditions at va...

his paper studies the symbolic reachability relations of a class of parameterized systems in the framework of regular model checking. The modules of each system are instantiated from a globally synchronized template, and each globally synchronized template is represented by a finite state automaton whose event set consists of global events and loca...

This paper presents a distributed implementation for a genetic algorithm, using Apache Spark, a fast and popular data processing framework. Our approach is rather general, but in this paper the parallelized genetic algorithm is used for test data generation for executable programs. The viability of the approach is demonstrated on two examples.

We present a binary analysis framework based on symbolic execution with the distinguishing capability to execute stepwise forward and also backward through the execution tree. It was developed internally at Bitdefender and code-named RIVER. The framework provides components such as a taint engine, a dynamic symbolic execution engine, and integratio...

We present a binary analysis framework based on symbolic execution with the distinguishing capability to execute stepwise forward and also backward through the execution tree. It was developed internally at Bitdefender and code-named RIVER. The framework provides components such as a taint engine, a dynamic symbolic execution engine, and integratio...

It is shown that the problem whether an arbitrary regular language has a non-empty decomposable sublanguage with respect to a fixed distribution is decidable if and only if the independence relation induced by the distribution is transitive. A sufficient condition on the distributed control architecture is then derived, under which there exist some...

Mobile distributed systems involve specific aspects such as migration, communication and concurrency, usually under temporal constraints. In this paper, we deal with formal modelling of timed migrating and communicating processes, as provided by the TIMO calculus. In this framework, mobile processes can move between different locations and communic...

This note studies two closely related decision problems from the area of decentralized synthesis. The first problem is about deciding the existence of a non-empty decomposable sublanguage of a regular language, called Decomposable Sublanguage Problem; the second is Joint Observability Problem, which is known in the literature. We provide characteri...

We propose an approach which, given a state-transition model of a system, constructs, in parallel, an approximate automaton model and a test suite for the system. The approximate model construction relies on a variant of Angluin's automata learning algorithm, adapted to finite cover automata. A finite cover automaton represents an approximation of...

Service-based applications are based on modern architectures that require careful design of interfaces and protocols to allow smooth integration of service components. These design artifacts are not only useful for implementation, but could also be used for the derivation of integration tests. In order to be applied in these different activities of...

Kernel P systems have been introduced with the aim of unifying different variants of P systems. They provide a coherent framework for specifying and solving a broad spectrum of problems. The tools built so far have aimed to formally verify systems modelled as kernel P systems, but the simulators obtained are not very efficient. In this paper we sho...

In this chapter we present some results obtained in the study of the bacterium E. coli related to its behavior at different level of oxygen in the environment. The biological model is expressed in terms of different molecules and their reactions. First, an agent-based model of E. coli is implemented in the FLAME framework for multi-agents and some...

Zielonka's theorem, established 25 years ago, states that any regular language closed under commutation is the language of an asynchronous automaton (a tuple of automata, one per process, exchanging information when performing common actions). Since then, constructing asynchronous automata has been simplified and improved [6, 19, 7, 12, 8, 4, 2, 20...

There have been few experience reports from industry on how Model-Driven Engineering (MDE) is applied and what the benefits are. This paper summarizes the experiences of three large industrial participants in a European research project with the objective of developing techniques and tools for applying MDE on the development of large and complex so...

Enterprise software helps modern corporates to automate their businesses in order to run efficiently and economically. We report a story of successful introduction of formal methods to business software development. This deployment came in three phases: modelling, formal verification and model-based testing. For each phase, we describe a few repres...

A computer-implemented method of evaluating interactions between processes. The method includes storing a message choreography meta-model that includes an interaction meta class and a constraint meta-class. A user creates a message choreography model between a first process and a second process. The method further includes executing a verification...

Model-Driven Engineering (MDE) has been promoted for many years as a means for handling the complexity of software development by raising the abstraction level and automating labor-intensive and error-prone tasks. However, there is little empirical evidence of the acceptance of MDE in industry which is the subject of this paper. The goal of this em...

This paper shows that language based top-down supervisor synthesis of Ramadge-Wonham supervisory control theory is in general not feasible. We show this as a direct consequence of the undecidability result of Decomposable Subset problem (and its prefix closed version), which in turn is a corollary of the undecidability result of Trace Closed Subset...

Event-B is a formal method for reliable systems specification and verification, which uses model refinement and decomposition as techniques to scale the design of complex systems. In previous work, we proposed an iterative approach for test generation and state model inference based on a variant of Angluin's learning algorithm, which integrates wel...

The Event-B method is a formal approach for reliable systems specification and verification, being supported by the Rodin platform, which includes mature plugins for theorem-proving, model-checking, or model (de)composition features. In order to complement these techniques with test generation and state model inference from Event-B models, we devel...

We present a computational framework for automatic synthesis of control and communication strategies for a robotic team from task specifications that are given as regular expressions about servicing requests in an environment. We assume that the location of the requests in the environment and the robot capacities and cooperation requirements to ser...

Event-B is a formal modeling language having set theory as its mathematical foundation and abstract state machines as its behavioral specifications. The language has very good tool support based on theorem proving and model checking technologies, but very little support for test generation. Motivated by industrial interest in the latter domain, thi...

This position paper discusses the challenges and opportunities of applying search-based techniques to a for- mal environment of abstract state machines defined using a language called Event-B. Event-B is based on a formal abstract machine notation that has a mature tool support and gets continuous feedback from industry. Although search- based tech...

We consider the following problem: GIVEN (1) a set of service requests occurring at known locations in an environment, (2) a set of temporal and logical constraints on how the requests need to be serviced, (3) a team of robots and their capacities to service the requests individually or through collaboration, FIND robot control and communication st...

This paper presents a case study for the modeling and model-based testing (MBT) of enterprise service choreographies. Our proposed MBT approach uses proprietary models called Message Choreography Models (MCM) as test models. The case study illustrates how MCM-based service integration testing allows to formalize design decisions and enables full in...

In this paper we present an approach of using model-driven technologies for testing of service component interactions. We report on an industrial experiment with a novel combination of existing UML standards, i.e., the UML Testing Profile (U2TP), in conjunction with proprietary domain specific languages (DSLs). Many model-based testing (MBT) approa...

Software development and testing of Enterprise Resource Planning (ERP) systems demands dedicated methods to tackle its special features. As manual testing is not able to systematically test ERP systems due to the involved complexity, an efficient testing approach should be automatic. Since the underlying business processes of enterprise systems are...

We present a computational framework for automatic synthesis of con-trol and communication strategies for a robotic team from task specifications given as regular expressions about servicing requests in an environment. Our approach is based on two main ideas. First, we extend recent results from formal synthesis of distributed systems to check for...

The concepts of Service Oriented Architectures (SOA) have manifested themselves as the next evolutionary step for software development and the industry is fast in adopting SOA. The gained freedom of combining loosely coupled services is countered by resulting challenges for testing. While current research is mainly focusing on the investigation of...

The success of service-oriented architectures (SOA) depends on faultless and seamless service integration. Formal modeling of global communication protocols between services enables a model-based integration testing (MBIT) approach. In this paper we present an MBIT approach based on SAP proprietary choreography models called message choreography mo...

Component integration plays a decisive role in service-oriented architectures (SOAs). The technical implementation must faithfully reflect business and enterprise integration requirements. This implies a good understanding of the globally observable message choreography but also of how messages are handled by the involved components and by the SOA...

Service choreographies describe the global communication protocols between services and testing these choreographies is an important task in the context of service-oriented architectures (SOA). Formal modeling of service choreographies makes a model-based testing (MBT) approach feasible. In this paper we present an MBT approach for SOA integration...

Service-oriented architecture (SOA) enables organizations to transform their existing IT infrastructure into a more flexible business process platform.. In this architecture, decoupled components that provide standard services can be composed to form individually configured and highly flexible applications. When building such applications it is imp...

This paper suggests a holistic design and development method combining test-driven and model-driven development for SOA architectures. It uses test-driven development on component level and model-based testing on system level. Moreover, monitored performance parameters during test execution serve as input for a model-driven performance analysis of...

The execution of a reactive system amounts to the repetitions of executions of control flow cycles in the component processes of the system. The way in which cycle executions are combined is not arbitrary since cycles may depend on or exclude one another. We believe that the information of such dependencies is important to the design, understanding...

Software development and testing of enterprise resource planning (ERP) systems demands dedicated methods to tackle its special features. As manual testing is not able to systematically test ERP systems due to the involved complexity, an effective testing approach should be automated, also requiring that the appropriate test data has to be provided...

This thesis investigates the synthesis problem for two classes of distributed transition systems: synchronous products and asynchronous automata. The underlying structure of these models consist of local automata synchronizing on common actions. The synthesis problem discussed is as follows: Given a global specification as a transition system TS an...

We describe an incomplete but sound and efficient livelock freedom test for infinite state asynchronous reactive systems. The method abstracts a system into a set of simple control flow cycles labeled with their message passing effects. From these cycles, it constructs a homogeneous integer programming problem (IP) encoding a necessary condition fo...

We consider the distributed implementability problem: Given a labeled transition system TS together with a distribution Δ of its actions over a set of processes, does there exist a distributed system over Δ such that its global transition system is 'equivalent' to TS? We work with the distributed system models of synchronous products of transition...