publications

2025

1. Beat the Offers—A Machine-Learning Approach for Predicting Contestants’ Choices and Games’ Outcomes on a TV Quiz Show

   Hana Ivandic, Branimir Pervan, Josip Knezovic, and 1 more author

   Applied Sciences, May 2025

   Abs DOI Bib

   Beat the Chasers is a popular UK-originating TV quiz show that premiered in Croatia in 2023. On the show, a contestant challenges a team of up to five chasers with respect to the offers provided by the production. Each offer balances risk and reward, varying in prize money, time advantage, and the number of chasers. In this paper, we first present the dataset obtained by extracting data from the publicly broadcast episodes of Beat the Chasers in Croatia. We then apply various machine-learning models with the goals of predicting (1) which offer a contestant is most likely to select and (2) the game’s outcome. The best-case results suggest that we can successfully do both by reaching an F1-score of 73.6% for the selected offer prediction and 84.6% for the game’s outcome prediction. Regarding the feature importance analysis, we identified the contestant’s hometown size, NUTS 2 region, age group, and gender as the most relevant features in the case of the selected offer prediction. As for the outcome prediction, the game-specific features emerged as the most important, namely, the cash builder result, the selected number of chasers, and the chasers’ time in the selected offer.

   @article{Ivandic2025,
     author = {Ivandic, Hana and Pervan, Branimir and Knezovic, Josip and Jovic, Alan},
     doi = {10.3390/app15105722},
     issn = {2076-3417},
     issue = {10},
     journal = {Applied Sciences},
     keywords = {Beat the Chasers,TV quiz show,data extraction,data mining,machine learning},
     month = may,
     pages = {5722},
     title = {{Beat the Offers—A Machine-Learning Approach for Predicting Contestants’ Choices and Games’ Outcomes on a TV Quiz Show}},
     volume = {15},
     url = {https://www.mdpi.com/2076-3417/15/10/5722},
     year = {2025}
   }

2023

1. MIDOM—A DICOM-Based Medical Image Communication System

   Branimir Pervan, Sinisa Tomic, Hana Ivandic, and 1 more author

   Applied Sciences, May 2023

   Abs DOI Bib HTML

   Despite the existing medical infrastructure being limited in terms of interoperability, the amount of medical multimedia transferred over the network and shared through various channels increases rapidly. In search of consultations with colleagues, medical professionals with the consent of their patients, usually exchange medical multimedia, mainly in the form of images, by using standard instant messaging services which utilize lossy compression algorithms. That consultation paradigm can easily lead to losses in image representation that can be misinterpreted and lead to the wrong diagnosis. This paper presents MIDOM—Medical Imaging and Diagnostics on the Move, a DICOM-based medical image communication system enhanced with a couple of variants of our previously developed custom lossless Classification and Blending Predictor Coder (CBPC) compression method. The system generally exploits the idea that end devices used by the general population and medical professionals alike are satisfactorily performant and energy-efficient, up to a point to support custom and complex compression methods successfully. The system has been implemented and appropriately integrated with Orthanc, a lightweight DICOM server, and a medical images storing PACS server. We benchmarked the system thoroughly with five real-world anonymized medical image sets in terms of compression ratios and latency reduction, aiming to simulate scenarios in which the availability of the medical services might be hardly reachable or in other ways limited. The results clearly show that our system enhanced with the compression methods in the question pays off in nearly every testing scenario by lowering the network latency to at least 60% of the latency required to send raw and uncompressed image sets and 25% in the best-case, while maintaining the perfect reconstruction of medical images and, thus, providing a more suitable environment for healthcare applications.

   @article{Pervan2023,
     author = {Pervan, Branimir and Tomic, Sinisa and Ivandic, Hana and Knezovic, Josip},
     doi = {10.3390/app13106075},
     journal = {Applied Sciences},
     number = {10},
     pages = {1--19},
     publisher = {MDPI},
     title = {{MIDOM—A DICOM-Based Medical Image Communication System}},
     url = {https://www.mdpi.com/2076-3417/13/10/6075},
     volume = {13},
     year = {2023},
   }

2022

1. Energy-efficient distributed password hash computation on heterogeneous embedded system

   Branimir Pervan, Josip Knezovic, and Emanuel Guberovic

   Automatika, May 2022

   Abs DOI Bib HTML

   This paper presents the improved version of our cool Cracker cluster (cCc), a heterogeneous distributed system for parallel and energy-efficient bcrypt password hash computation. The cluster consists of up to 8 computational units (nodes) with different performances measured in bcrypt hash computations per second [H/s]. In the cluster, nodes are low-power heterogeneous embedded systems with programmable logic containing specialized hash computation accelerators. In the experiments, we used a combination of Xilinx Zynq-series SoC boards and ZTEX 1.15y board which was initially used as a bitcoin miner. Zynq based nodes use the improved version of our custom bcrypt accelerator, which executes the most costly parts of the bcrypt hash computation in programmable logic. The cluster was formed around the famous open-source password cracking software package John the Ripper (abbr. JtR). On the communication layer, we used Message Passing Interface (MPI)library with a standard Ethernet network connecting the nodes. To mitigate the different performances among the cluster nodes and to balance the load, we developed and implemented password candidate distribution scheme based on the passwords’ probability distribution, i.e. the order of appearance in the dictionary. We tested individual nodes and the cluster as a whole, trying different combinations of nodes and evaluating our distribution scheme for password candidates. We also compared our cluster with various GPU implementations in terms of performance, energy-efficiency, and price-efficiency. We show that our solution outperforms other platforms such as high-end GPUs, by a factor of at least 3 in terms of energy-efficiency and thus producing less overall cost of password attack than other platforms. In terms of the total operational costs, our cluster pays off after 4500 cracked passwords for a bcrypt hash with cost parameter 12, which makes it more appealing for real-world password-based system attacks. We also demonstrate the scalability of our cCc cluster.

   @article{Pervan2022,
     author = {Pervan, Branimir and Knezovic, Josip and Guberovic, Emanuel},
     doi = {10.1080/00051144.2022.2042115},
     journal = {Automatika},
     number = {3},
     pages = {399--417},
     publisher = {Taylor & Francis},
     title = {{Energy-efficient distributed password hash computation on heterogeneous embedded system}},
     url = {https://www.tandfonline.com/doi/abs/10.1080/00051144.2022.2042115},
     volume = {63},
     year = {2022},
   }

2020

1. A Survey on Parallel Architectures and Programming Models

   Branimir Pervan and Josip Knezovic

   In 2020 43rd International Convention on Information, Communication and Electronic Technology (MIPRO), Sep 2020

   Abs DOI Bib HTML

   In addition to large-scale computers, multicore processors have taken a significant part in all kinds of devices, from personal computers to cell phones. Although programming techniques for parallel systems exist for a while, the development of applications that can appropriately utilize multicores is still challenging in many aspects, especially for full exploitation of the computational resources. Additionally, another challenge is the efficient and easy programming of heterogeneous systems for the complete exploitation of silicon resources. Solutions to making parallel programming more developer-friendly are various programming models that abstract parallelism and concurrency. Implementations of those models need to extend even to lower layers of software parallelism and hardware parallelism as well.This paper gives an overview of parallel architectures and trending programming models for such processing units and systems. It also presents challenges to scalability and portability in parallel systems and presents up to date trends in heterogeneous systems that heavily exploit parallelism.

   @inproceedings{Pervan2020,
     author = {Pervan, Branimir and Knezovic, Josip},
     booktitle = {2020 43rd International Convention on Information, Communication and Electronic Technology (MIPRO)},
     doi = {10.23919/MIPRO48935.2020.9245341},
     isbn = {978-953-233-099-1},
     keywords = {Domain-specific languages,Heterogeneous systems,Parallel architectures,Parallel programming,Portability,Reactive programming,Scalability},
     month = sep,
     pages = {999--1005},
     publisher = {IEEE},
     title = {{A Survey on Parallel Architectures and Programming Models}},
     url = {https://ieeexplore.ieee.org/document/9245341/},
     year = {2020},
   }

2019

1. Distributed Password Hash Computation on Commodity Heterogeneous Programmable Platforms

   Branimir Pervan, Josip Knezovic, and Katja Pericin

   In 13th USENIX Workshop on Offensive Technologies, WOOT 2019, co-located with USENIX Security 2019, Sep 2019

   Abs Bib HTML

   In this paper, we present the Cool Cracker Cluster cCc: a heterogeneous distributed system for parallel, energy-efficient, and high-speed bcrypt password hash computation. The cluster consists of up to 32 heterogeneous nodes with Zynq-7000-based SoCs featuring a dual-core, general-purpose ARM processor coupled with FPGA programmable logic. Each node uses our custom bcrypt accelerator which executes the most costly parts of the hash computation in programmable logic. We integrated our bcrypt implementation into John the Ripper, an open source password cracking software. Message Passing interface (MPI) support in John the Ripper is used to form a distributed cluster. We tested the cluster, trying different configurations of boards (Zedboards and Pynq boards), salt randomness, and cost parameters finding out that password cracking scales linearly with the number of nodes. In terms of performance (number of computed hashes per second) and energy efficiency (performance per Watt), cCc outperforms current systems based on high-end GPU cards, namely Nvidia Tesla V100, by a factor of 2.72 and 5 respectively.

   @inproceedings{Pervan2019,
     author = {Pervan, Branimir and Knezovic, Josip and Pericin, Katja},
     booktitle = {13th USENIX Workshop on Offensive Technologies, WOOT 2019, co-located with USENIX Security 2019},
     title = {{Distributed Password Hash Computation on Commodity Heterogeneous Programmable Platforms}},
     url = {https://www.usenix.org/conference/woot19/presentation/pervan https://www.usenix.org/system/files/woot19-paper_pervan.pdf},
     year = {2019},
   }

2. Deep Learning Accelerator on Programmable Heterogeneous System with RISC-V Processor

   Luka Strizic, Branimir Pervan, and Josip Knezovic

   In 2019 proceedings of the 42nd international convention MIPRO, Sep 2019

   Abs Bib HTML

   This paper explores a heterogeneous, open source, RISC-V based platform, called PULP (Parallel Ultra-Low-Power Processing Platform) with customizable MAC (multiply and accumulate) accelerator for neural network acceleration on edge computing devices with low power and performance-limited resources. To address the performance and power constraints we propose binarized implementation of a neural network accelerator. We show that binarized implementation is faster than the fixed-point accelerated implementation, albeit with some loss in precision, but still applicable to most of the use cases in edge computing.

   @inproceedings{Strizic2019,
     author = {Strizic, Luka and Pervan, Branimir and Knezovic, Josip},
     booktitle = {2019 proceedings of the 42nd international convention MIPRO},
     pages = {1126--1131},
     publisher = {IEEE},
     title = {{Deep Learning Accelerator on Programmable Heterogeneous System with RISC-V Processor}},
     url = {https://www.bib.irb.hr/1021661/download/1021661.16_cts_5493.pdf},
     year = {2019},
   }

3. Hazelnut - An Energy Efficient Base IoT Module for Wide Variety of Sensing Applications

   Branimir Pervan, Emanuel Guberovic, and Filip Turcinovic

   In Proceedings of the 6th Conference on the Engineering of Computer Based Systems, Sep 2019

   Abs DOI Bib HTML

   With the everlasting expansion of Internet of Things (IoT), backed by the increased availability of cheap hardware making it widely available, more and more modules are finding their usage in everyday situations. In order to maximize the pervasiveness of IoT modules, human intervention must be reduced to a bare minimum. This primarily addresses the need for relatively frequent battery charging, since cheap microcontrollers for IoT modules available to the masses are still significantly energy inefficient. This renders such modules hardly usable in amateur and semi-professional environments. As a solution to the problem of cheap but energy inefficient modules, this paper introduces a concept of designing an IoT module of two microcontrollers: one powerful and peripheral-rich with networking capabilities, and one extremely energy-saving, used to wake up the first one from a deep sleep mode when richer set of functionalities is needed or more intensive tasks are processed. The system is described from a top view along with its implementation in a case study. Power consumption is evaluated practically through a well-defined measurement methodology and comparison of the results. We show that by adding an additional cheap and energy-saving microcontroller, one can achieve significant energy savings with an insignificant rise in the overall price of the system.

   @inproceedings{Pervan2021,
     address = {New York, NY, USA},
     author = {Pervan, Branimir and Guberovic, Emanuel and Turcinovic, Filip},
     booktitle = {Proceedings of the 6th Conference on the Engineering of Computer Based Systems},
     doi = {10.1145/3352700.3352702},
     isbn = {9781450376365},
     issn = {21531633},
     keywords = {Embedded systems,Energy efficiency,Energy saving,Internet of things},
     month = sep,
     pages = {1--4},
     publisher = {ACM},
     title = {{Hazelnut - An Energy Efficient Base IoT Module for Wide Variety of Sensing Applications}},
     url = {https://dl.acm.org/doi/10.1145/3352700.3352702},
     year = {2019},
   }

4. Project Houseleek - A Case Study of Applied Object Recognition Models in Internet of Things

   Jure Knezovic, Branimir Pervan, Zvonimir Relja, and 1 more author

   In 2019 proceedings of the 42nd international convention MIPRO, Sep 2019

   Abs Bib HTML

   Nowadays, the gap between academic work and practical application of that work is rapidly diminishing. This fact can be backed by several factors: the increase in availability of the research results, as well as research artifacts; the rise in the level of education in general; the availability of broadband networks and the more affordable prices of the technology used for research. Also, due to the pervasion of the technology in all spheres of society, there is an emergence of new possibilities of applying disruptive technologies at all levels, including homes or workplaces of individual users. This paper presents Project Houseleek: a multilayer system that utilizes disruptive technologies to enhance and facilitate access to individual premises in smart areas. On the authentication layer, the system uses disruptive deep learning technologies to identify or learn itself a person in a real-world environment from an image grabbed in relatively rough conditions, while at the authorization layer it learns at runtime the access rights to specific parts of the smart area for that person. The testing system is implemented at the Department of Control and Computer Engineering, Faculty of Electrical Engineering and Computing where it exceeded the expectations of the users on both authentication and authorization layers.

   @inproceedings{Knezovic2019,
     author = {Knezovic, Jure and Pervan, Branimir and Relja, Zvonimir and Knezovic, Josip},
     booktitle = {2019 proceedings of the 42nd international convention MIPRO},
     number = {May},
     pages = {1051--1055},
     publisher = {IEEE},
     title = {{Project Houseleek - A Case Study of Applied Object Recognition Models in Internet of Things}},
     url = {https://www.bib.irb.hr/1021658/download/1021658.04_cts_5327.pdf},
     year = {2019},
   }