In this paper, we presents a low-complexity deep learning frameworks for acoustic scene classification (ASC). The proposed framework can be separated into three main steps: Front-end spectrogram extraction, back-end classification, and late fusion of predicted probabilities. First, we use Mel filter, Gammatone filter and Constant Q Transfrom (CQT) to transform raw audio signal into spectrograms, where both frequency and temporal features are presented. Three spectrograms are then fed into three individual back-end convolutional neural networks (CNNs), classifying into ten urban scenes. Finally, a late fusion of three predicted probabilities obtained from three CNNs is conducted to achieve the final classification result. To reduce the complexity of our proposed CNN network, we apply two model compression techniques: model restriction and decomposed convolution. Our extensive experiments, which are conducted on DCASE 2021 (IEEE AASP Challenge on Detection and Classification of Acoustic Scenes and Events) Task 1A development dataset, achieve a low-complexity CNN based framework with 128 KB trainable parameters and the best classification accuracy of 66.7%, improving DCASE baseline by 19.0%

The European Union's Internal Security Fund (ISF) will contribute to ensuring a high level of security in the Union, by supporting actions that help to prevent and combat terrorism and radicalisation, serious and organised crime, and cybercrime. One such project that has launched in January 2022 is Anti-FinTer: Versatile artificial intelligence investigative technologies for revealing online cross-border financing activities of terrorism.

In this paper, we present deep learning frameworks for audio-visual scene classification (SC) and indicate how individual visual and audio features as well as their combination affect SC performance.Our extensive experiments, which are conducted on DCASE (IEEE AASP Challenge on Detection and Classification of Acoustic Scenes and Events) Task 1B development dataset, achieve the best classification accuracy of 82.2\%, 91.1\%, and 93.9\% with audio input only, visual input only, and both audio-visual input, respectively.The highest classification accuracy of 93.9\%, obtained from an ensemble of audio-based and visual-based frameworks, shows an improvement of 16.5\% compared with DCASE baseline.

This report shows a deep learning framework for audio-visual scene classification (SC). Our extensive experiments, which are conducted on DCASE Task 1B development dataset, achieve the best classification accuracy of 82.2%, 91.1%, and 93.9% with audio input only, visual input only, and both audiovisual input, respectively.

There is currently an increasing demand for cryptoasset analysis tools among cryptoasset service providers, the financial industry in general, as well as across academic fields. At the moment, one can choose between commercial services or low-level open-source tools providing programmatic access. In this paper, we present the design and implementation of another option: the GraphSense Cryptoasset Analytics Platform, which can be used for interactive investigations of monetary flows and, more importantly, for executing advanced analytics tasks using a standard data science tool stack. By providing a growing set of open-source components, GraphSense could ultimately become an instrument for scientific investigations in academia and a possible response to emerging compliance and regulation challenges for businesses and organizations dealing with cryptoassets.

This paper presents an inception-based deep neural network for detecting lung diseases using respiratory sound input. Recordings of respiratory sound collected from patients are first transformed into spectrograms where both spectral and temporal information are well represented, in a process referred to as front-end feature extraction. These spectrograms are then fed into the proposed network, in a process referred to as back-end classification, for detecting whether patients suffer from lung-related diseases. Our experiments, conducted over the ICBHI benchmark metadataset of respiratory sound, achieve competitive ICBHI scores of 0.53/0.45 and 0.87/0.85 regarding respiratory anomaly and disease detection, respectively.

The architectural design of our energy systems dates back to a time without information technology (IT). Over time, IT was applied where it increased efficiency and safety. About 12 years ago, the Smart Grid era began. In the meantime, we talk about digitalization. Electrical energy systems require embedded systems, Internet of Things, computation clusters and data analytics. However, IT also has another role in the energy system, namely that of a substantial consumer. Crypto currencies and data centers are on the rise. We analyze impacts on energy demand and discuss risks and chances of this development.

The forensic investigation of a terrorist attack poses a significant challenge to the investigative authorities, as often several thousand hours of video footage must be viewed. Large scale Video Analytic Platforms (VAP) assist law enforcement agencies (LEA) in identifying suspects and securing evidence. Current platforms focus primarily on the integration of different computer vision methods and thus are restricted to a single modality. We present a video analytic platform that integrates visual and audio analytic modules and fuses information from surveillance cameras and video uploads from eyewitnesses. Videos are analyzed according their acoustic and visual content. Specifically, Audio Event Detection is applied to index the content according to attack-specific acoustic concepts. Audio similarity search is utilized to identify similar video sequences recorded from different perspectives. Visual object detection and tracking are used to index the content according to relevant concepts. Innovative user-interface concepts are introduced to harness the full potential of the heterogeneous results of the analytical modules, allowing investigators to more quickly follow-up on leads and eyewitness reports.

The Data Market Austria (DMA) is an ecosystem of federated data and service infrastructures. It aims at establishing a market platform where data assets can be made accessible and offered for purchase. To support this, the DMA offers a central portal with catalogue and search services as well as a set of microservices for metadata mapping or enrichment, data quality assessment, data set submission, storage, management, and dissemination. The DMA relies on blockchain technology to allow a network of DMA members sharing information about the provenance and trading of datasets and services. This paper describes the blockchain application scenarios and the implementation of the blockchain-based distributed setup of the DMA.

With the booming development of smartphone capabilities, these devices are increasingly frequent victims of targeted attacks in the ‘silent battle’ of cyberspace. Protecting Android smartphones against the increasing number of malware applications has become as crucial as it is complex. To be effective in identifying and defeating malware applications, cyber analysts require novel distributed detection and reaction methodologies based on information security techniques that can automatically analyse new applications and share analysis results between smartphone users. Our goal is to provide a real-time solution that can extract application features and find related correlations within an aggregated knowledge base in a fast and scalable way, and to automate the classification of Android smartphone applications. Our effective and fast application analysis method is based on artificial intelligence and can support smartphone users in malware detection and allow them to quickly adopt suitable countermeasures following malware detection. In this paper, we evaluate a deep neural network supported by word-embedding technology as a system for malware application classification and assess its accuracy and performance. This approach should reduce the number of infected smartphones and increase smartphone security. We demonstrate how the presented techniques can be applied to support smartphone application classification tasks performed by smartphone users.

Protecting Critical Infrastructure (CI) against increasing cyber threats has become as crucial as it is complicated. To be effective in identifying and defeating cyber attacks, cyber analysts require novel distributed detection and reaction methodologies based on information security techniques that can automatically analyse incident reports and securely share analysis results between Critical Infrastructure stakeholders. Our goal is to provide solutions in real-time that could replace human input for cyber incident analysis tasks (triage) to classify cyber incident reports, find related reports in a fast and scalable way, eliminate irrelevant information, and automate reporting life- cycle management. Our effective and fast incident management method is based on artificial intelligence and can support cyber analysts in establishing cyber situational awareness, and allow them to quickly adopt suitable countermeasures in the case of an attack. In this paper, we evaluate deep autoencoder neural network supported by Blockchain technology as a system for incident classification and management, and assess its accuracy and performance. This approach should reduce the number of manual operations and save storage space. We used a Blockchain smart contract technique to provide an automated trusted system for incident management workflow that allows automatic acquisition, classification and enrichment of incident data. We demonstrate how the presented techniques can be applied to support incident handling tasks performed by security operation centres.

Nowadays, cyber critical infrastructures (CIs) are increasingly targeted by highly sophisticated cyber attacks and should be protected. Advances in cyber situational awareness technology lead to the creation of increasingly complex tools. Human analysts face challenges finding relevant information in large, complex data sets, when exploring data to discover patterns and insights. To be effective in identifying and defeating future cyber-attacks, cyber analysts require novel tools for incident report classification and life cycle management that can automatically analyse and share result in secure way between CI stakeholders to achieve better situation comprehension. Our goal is to provide solutions in realtime that could replace human input for cyber incident classification and management tasks to eliminate irrelevant information and to focus on important information to promptly adopt suitable countermeasures in case of an attack. Another contribution relates to the provided support for document life cycle management that should reduce the number of manual operations and save storage space. In this paper we evaluate the application of so-called “smart contracts” to an incident classification system and assess its accuracy and performance. We demonstrate how the presented techniques can be applied to support incident handling tasks performed by security operation centers (SOCs).

Modern critical infrastructures are increasingly targeted by highly sophisticated cyber attacks and are protected by increasingly complex tools. Cyber analysts face many challenges finding relevant information in large, complex data sets, and require novel distributed detection and reaction methodologies based on secured transaction techniques. These technologies should automatically analyse incident report and share analysis result in secure way between critical infrastructure stakeholders to achieve better situational awareness. Our goal is to provide solutions in real-time that could replace human input for cyber incident analysis tasks (Triage) to remove false positives and to eliminate irrelevant information. The effective and fast warning system should support cyber analyst to establish cyber situational awareness, and allow analysts to promptly respond in case of an attack. In this paper we evaluate the application of so-called “smart contracts” to an incident warning system and assess its accuracy and performance. We demonstrate how the presented techniques can be applied to support incident handling tasks performed by security operation centers. We show that a real-time “smart contracts” solution can replace human input for a large number of threat intelligence analysis tasks.

Digitization workflows for automatic acquisition of image collections are susceptible to errors and require quality assurance. This paper presents the automated quality assurance tools aiming at detection of possible quality issues that supports decision making for document image collections. The main contribution of this research is the implementation of various image processing tools for different error detection scenarios and their combination in to a single tool suite. The tool suite includes: (1) The matchbox tool for accurate near-duplicate detection in document image collections, based on SIFT feature extraction. (2) The finger detection tool aims at automatic detection of fingers that mistakenly appear in scans from digitized image collections, which uses processing techniques for edge detection, local image information extraction and its analysis for reasoning on scan quality. (3) The cropping error detection tool supports the detection of common cropping problems such as text shifted to the edge of the image, unwanted page borders, or unwanted text from a previous page on the image. Another important contribution of this work is a definition of the quality assurance workflow and its automatic execution for error detection in digital document collections. The presented tool suite detects described errors and presents them for additional manual analysis and collection cleaning. A statistical overview of evaluated data and characteristics like performance and accuracy is delivered. The results of the analysis confirm our hypothesis that an automated approach is able to detect errors with reliable quality, thus making quality control for large digitisation projects a feasible and affordable process.

One of the common challenges in the mass-digitisation of book collections is correctly cropping (removing unnecessary border material from the digital image) during the automated image post-processing. This paper presents a method that supports the analysis of digital collections (e.g. JPG files) for detecting common cropping problems such as text shifted to the edge of the image, unwanted page borders, or unwanted text from a previous page on the image. One contribution of this work is a definition of the evaluation use cases for cropping problems. A second contribution is the creation of a reliable expert tool for document cropping error detection based on image profiling techniques. This tool can be applied in quality assurance workflows for digital book collections. Our suggested method employs evaluation parameters that can be defined for each book. The tool works independently of the image size, format and colour. We have analysed two real world collections with correct and corrupted images, and our tool has demonstrated good recall and precision for both corrupted image and correct images.

This paper presents a system for Braille learning support using real-time panoramic views generated from the novel smart panorama camera 360SCAN. The system makes use of the modern image processing libraries and state-of-the-art features extraction and clustering methods. We compare the real-time frames recorded by the bio-inspired camera to the reference images in order to determine particular figures. One contribution of the proposed method is that image edges can be transformed to the presentation on Braille display directly without any image processing. It is possible due to the bio-inspired construction of camera sensor. Another contribution is that our approach provides Braille users with images recorded from natural scenes. We conducted several experiments that verify the methods that demonstrate learning figures captured by the smart camera. Our goal is to process such images and present them on the Braille Display in a form appropriate for visually impaired people. All evaluations were performed in the natural environment with ambient illumination of 200 lux, which demonstrates high camera reliability in difficult light conditions. The system can be optimized by applying additional filters and features algorithms and by decreasing the rotational speed of the camera. The presented Braille learning support system is a building block for a rich and qualitative educational system for the efficient information transfer focused on visually impaired people.

While interoperability between “live” e-government systems has been a major work area during the last decade, the fact that much of this information needs to be preserved for the long-term after the initial creation, and reuse has been of secondary concern. This paper looks into the needs of long-term preservation of digitalborn e-government data and proposes further actions to address the challenge in a cost-effective manner. General Terms infrastructure, communities, strategic environment, digital preservation marketplace, case studies and best practice, training and education.

This paper presents a system for quality control of real-time panoramic views generated from the novel smart panorama camera 360SCAN. The system makes use of the modern image processing library OpenIMAJ and state-of-the-art features extraction and clustering methods. We compare a real-time frame collection recorded by the camera to a reference image collection in order to determine camera readiness. We conducted several experiments that verify the methods that demonstrate smart camera operational status and evaluate changes in the position or number of objects in the working location. All evaluations were performed in the natural environment with ambient illumination of 200 lux, which demonstrates high camera reliability in difficult light conditions. The system can be optimized for embedded applications by applying additional filters and features algorithms and by decreasing the rotational speed of the camera. The presented quality control system is a building block for a rich and qualitative expert system for the efficient control and support of the smart camera.

This paper presents an approach for automatic detection of fingers that mistakenly appear in scans from digitized image collections. Our goal is to create a reliable detection tool that is independent from scan quality, finger sizes, direction, shape, colour and lighting conditions. Modern image processing techniques are applied for edge detection, local image information extraction, and analysis. We employed expert knowledge to determine default parameters of the algorithm, and support customized parameters for specific institutional workflows. Results for three digital collections analysis are presented. Documents with finger artefacts are identified with high reliability and validated by human visual inspection. The proposed method achieves up to 86 percent classification accuracy.

The digital collections of scientific and memory institutions - many of which are already in the petabyte range - are growing larger every day. The fact that the volume of archived digital content worldwide is increasing geometrically, demands that their associated preservation activities become more scalable. The economics of long-term storage and access demand that they become more automated. The present state of the art fails to address the need for scalable automated solutions for tasks like the characterization or migration of very large volumes of digital content. Standard tools break down when faced with very large or complex digital objects; standard workflows break down when faced with a very large number of objects or heterogeneous collections. In short, digital preservation is becoming an application area of big data, and big data is itself revealing a number of significant preservation challenges.