Refine
Year of publication
Document Type
- Article (reviewed) (145)
- Master's Thesis (63)
- Conference Proceeding (60)
- Contribution to a Periodical (45)
- Bachelor Thesis (29)
- Working Paper (15)
- Study Thesis (13)
- Doctoral Thesis (5)
- Book (3)
- Part of a Book (1)
Conference Type
- Konferenzartikel (52)
- Konferenzband (4)
- Konferenz-Abstract (3)
Language
- English (382) (remove)
Has Fulltext
- yes (382) (remove)
Keywords
- COVID-19 (19)
- Government Measures (13)
- Export (11)
- Corona (9)
- 3D printing (8)
- Crisis (8)
- IT-Sicherheit (8)
- neuroprosthetics (6)
- Chromatography (5)
- Deep learning (5)
Institute
- Fakultät Maschinenbau und Verfahrenstechnik (M+V) (144)
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (77)
- Fakultät Wirtschaft (W) (53)
- INES - Institut für nachhaltige Energiesysteme (52)
- Fakultät Medien (M) (ab 22.04.2021) (45)
- Fakultät Elektrotechnik und Informationstechnik (E+I) (bis 03/2019) (32)
- IfTI - Institute for Trade and Innovation (26)
- ivESK - Institut für verlässliche Embedded Systems und Kommunikationselektronik (25)
- Fakultät Medien und Informationswesen (M+I) (bis 21.04.2021) (23)
- POIM - Peter Osypka Institute of Medical Engineering (16)
Open Access
- Open Access (293)
- Gold (66)
- Closed (45)
- Closed Access (44)
- Hybrid (34)
- Diamond (24)
- Bronze (23)
- Grün (7)
In a randomized controlled cross-over study ten male runners (26.7 ± 4.9 years; recent 5-km time: 18:37 ± 1:07 min:s) performed an incremental treadmill test (ITT) and a 3-km time trial (3-km TT) on a treadmill while wearing either carbon fiber insoles with downwards curvature or insoles made of butyl rubber (control condition) in light road racing shoes (Saucony Fastwitch 9). Oxygen uptake, respiratory exchange ratio, heart rate, blood lactate concentration, stride frequency, stride length and time to exhaustion were assessed during ITT. After ITT, all runners rated their perceived exertion, perceived shoe comfort and perceived shoe performance. Running time, heart rate, blood lactate levels, stride frequency and stride length were recorded during, and shoe comfort and shoe performance after, the 3-km TT. All parameters obtained during or after the ITT did not differ between the two conditions [range: p = 0.188 to 0.948 (alpha value: 0.05); Cohen's d = 0.021 to 0.479] despite the rating of shoe comfort showing better scores for the control insoles (p = 0.001; d = −1.646). All parameters during and after the 3-km TT showed no differences (p = 0.200 to 1.000; d = 0.000 to 0.501) between both conditions except for shoe comfort showing better scores for control insoles (p = 0.017; d = −0.919). Running with carbon fiber insoles with downwards curvature did not change running performance or any submaximal or maximal physiological or biomechanical parameter and perceived exertion compared to control condition. Shoe comfort is impaired while running with carbon fiber insoles. Wearing carbon fiber insoles with downwards curvature during treadmill running is not beneficial when compared to running with control insoles.
Garbage in, Garbage out: How does ambiguity in data affect state-of-the-art pedestrian detection?
(2024)
This thesis investigates the critical role of data quality in computer vision, particularly in the realm of pedestrian detection. The proliferation of deep learning methods has emphasised the importance of large datasets for model training, while the quality of these datasets is equally crucial. Ambiguity in annotations, arising from factors like mislabelling, inaccurate bounding box geometry and annotator disagreements, poses significant challenges to the reliability and robustness of the pedestrian detection models and their evaluation. This work aims to explore the effects of ambiguous data on model performance with a focus on identifying and separating ambiguous instances, employing an ambiguity measure utilizing annotator estimations of object visibility and identity. Through accurate experimentation and analysis, trade-offs between data cleanliness and representativeness, noise removal and retention of valuable data emerged, elucidating their impact on performance metrics like the log average miss-rate, recall and precision. Furthermore, a strong correlation between ambiguity and occlusion was discovered with higher ambiguity corresponding to greater occlusion prevalence. The EuroCity Persons dataset served as the primary dataset, revealing a significant proportion of ambiguous instances with approximately 8.6% ambiguity in the training dataset and 7.3% in the validation set. Results demonstrated that removing ambiguous data improves the log average miss-rate, particularly by reducing the false positive detections. Augmentation of the training data with samples from neighbouring classes enhanced the recall but diminished precision. Error correction of wrong false positives and false negatives significantly impacts model evaluation results, as evidenced by shifts in the ECP leaderboard rankings. By systematically addressing ambiguity, this thesis lays the foundation for enhancing the reliability of computer vision systems in real-world applications, motivating the prioritisation of developing robust strategies to identify, quantify and address ambiguity.
The interest of scientists to study motion sequences exists in the fields of sports science, clinical analysis and computer animation for quite some time. While in the last decades mainly markerbased motion capture systems have been used to evaluate movements, the interest in markerless systems is growing more and more. Nevertheless, in the field of clinical analysis, markerless methods have not yet proven their value, partly due to a lack of studies evaluating the quality of the obtained data. Therefore, this study aims to validate two markerless motion capture softwares from Simi Reality Motion Systems. The software Simi Shape, which is a mixture of traditional image-based tracking supported by an artificial intelligence net (AI net), and the software Crush, that uses a completely AI-based method. For this purpose, all motion data was recorded with two in-house motion capture systems. One system for recording the movements for a markerbased evaluation as gold standard and one system for markerless tracking. Within a laboratory environment, eight cameras per system were mounted around the area of motion. By placing two cameras in the same position and using the same calibration, deviations in the image data between those for markerbased and markerless tracking were extremely minimal. Based on this data, markerbased tracking was performed using the Simi Motion program, markerless tracking was performed using the Simi Shape software system and the latest software from Simi Reality Motion Systems, Crush. When comparing the markerless data with the markerbased data, an average root mean square error of 0,038 m was calculated for Simi Shape and a deviation of 0,037 m for Crush. In a direct comparison of the two markerless systems, a root mean square error of 0,019 m was scored. Based on these data, conclusions could be drawn about the accuracies of the two markerless systems. The obtained kinematic data of the tracking are in the range of high accuracy, which is limited to a deviation of less than 0,05 m according to the literature.
As the Industry 4.0 is evolving, the previously separated Operational Technology (OT) and Information Technology (IT) is converging. Connecting devices in the industrial setting to the Internet exposes these systems to a broader spectrum of cyber-attacks. The reason is that since OT does not have much security measures as much as IT, it is more vulnerable from the attacker's perspective. Another factor contributing to the vulnerability of OT is that, when it comes to cybersecurity, industries have focused on protecting information technology and less prioritizing the control systems. The consequences of a security breach in an OT system can be more adverse as it can lead to physical damage, industrial accidents and physical harm to human beings. Hence, for the OT networks, certificate-based authentication is implemented. This involves stages of managing credentials in their communication endpoints. In the previous works of ivESK, a solution was developed for managing credentials. This involves a CANopen-based physical demonstrator where the certificate management processes were developed. The extended feature set involving certificate management will be based on the existing solution. The thesis aims to significantly improve such a solution by addressing two key areas that is enhancing functionality and optimizing real-time performance. Regarding the first goal, firstly, an analysis of the existing feature set shall be carried out, where the correct functionality shall be guaranteed. The limitations from the previously implemented system will be addressed and to make sure it can be applied to real world scenarios, it will be implemented and tested in the physical demonstrator. This will lay a concrete foundation that these certificate management processes can be used in the industries in large-scale networks. Implementation of features like revocation mechanism for certificates, automated renewal of the credentials and authorization attribute checks for the certificate management will be implemented. Regarding the second goal, the impact of credential management processes on the ongoing CANopen real-time traffic shall be a studied. Since in real life scenarios, mission-critical applications like Industrial control systems, medical devices, and transportation networks rely on real-time communication for reliable operation, delays or disruptions caused by credential management processes can have severe consequences. Optimizing these processes is crucial for maintaining system integrity and safety. The effect to minimize the disturbance of the credential management processes on the normal operation of the CANopen network shall be characterized. This shall comprise testing real-time parameters in the network such as CPU load, network load and average delay. Results obtained from each of these tests will be studied.
Increasing global energy demand and the need to transition to sustainable energy sources to mitigate climate change, highlights the need for innovative approaches to improve the resilience and sustainability of power grids. This study focuses on addressing these challenges in the context of Morocco's evolving energy landscape, where increasing energy demand and efforts to integrate renewable energy require grid reinforcement strategies. Using renewable energy sources such as photovoltaic systems and energy storage technologies, this study aims to develop a methodology for strengthening rural community grids in Morocco.
Traditional reinforcement measures such as line and transformer upgrades will be investigated as well as the integration of power generation from photovoltaic systems, which offer a promising way to utilise Morocco's abundant solar resources. In addition, energy storage systems will be analysed as potential solutions to the challenges of grid stability and resilience. Using comprehensive data analysis, scenario planning and simulation methods with the open-source simulation software Panda Power, this study aims to assess the impact of different grid reinforcement measures, including conventional methods, photovoltaic integration, and the use of energy storage, on grid performance and sustainability. The results of this study provide valuable insights into the challenges and opportunities of transitioning to a more resilient and sustainable energy future in Morocco.
Based on a rural medium-voltage grid in Souihla, Morocco, three scenarios were carried out to assess the impact of demand growth in 2030 and 2040. The first scenario focuses on conventional grid reinforcement measures, while the second scenario incorporates energy from residential photovoltaic systems. The third scenario analyses the integration of storage systems and their impact on grid reinforcement in 2030.
The simulations with energy from photovoltaic systems show a reduction in grid reinforcement measures compared to the scenario without solar energy. In addition, the introduction of a storage system in 2030 led to a significant reduction in the required installed transformer capacity and fewer congested lines. Furthermore, the results emphasized the role of storage in stabilizing grid voltage levels.
In summary, the results highlighted the potential benefits of integrating energy from photovoltaics and storage into the grid. This integration not only reduces the need for transformers and overall grid infrastructure but also promotes a more efficient and sustainable energy system.
This paper provides a comprehensive overview of approaches to the determination of isocontours and isosurfaces from given data sets. Different algorithms are reported in the literature for this purpose, which originate from various application areas, such as computer graphics or medical imaging procedures. In all these applications, the challenge is to extract surfaces with a specific isovalue from a given characteristic, so called isosurfaces. These different application areas have given rise to solution approaches that all solve the problem of isocontouring in their own way. Based on the literature, the following four dominant methods can be identified: the marching cubes algorithms, the tessellation-based algorithms, the surface nets algorithms and the ray tracing algorithms. With regard to their application, it can be seen that the methods are mainly used in the fields of medical imaging, computer graphics and the visualization of simulation results. In our work, we provide a broad and compact overview of the common methods that are currently used in terms of isocontouring with respect to certain criteria and their individual limitations. In this context, we discuss the individual methods and identify possible future research directions in the field of isocontouring.
Background: Assistive Robotic Arms are designed to assist physically disabled people with daily activities. Existing joysticks and head controls are not applicable for severely disabled people such as people with Locked-in Syndrome. Therefore, eye tracking control is part of ongoing research. The related literature spans many disciplines, creating a heterogeneous field that makes it difficult to gain an overview.
Objectives: This work focuses on ARAs that are controlled by gaze and eye movements. By answering the research questions, this paper provides details on the design of the systems, a comparison of input modalities, methods for measuring the performance of these controls, and an outlook on research areas that gained interest in recent years.
Methods: This review was conducted as outlined in the PRISMA 2020 Statement. After identifying a wide range of approaches in use the authors decided to use the PRISMA-ScR extension for a scoping review to present the results. The identification process was carried out by screening three databases. After the screening process, a snowball search was conducted.
Results: 39 articles and 6 reviews were included in this article. Characteristics related to the system and study design were extracted and presented divided into three groups based on the use of eye tracking.
Conclusion: This paper aims to provide an overview for researchers new to the field by offering insight into eye tracking based robot controllers. We have identified open questions that need to be answered in order to provide people with severe motor function loss with systems that are highly useable and accessible.
"Ad fontes!"
Francesco Petrarca (1301–1374)
In the beginning, there was an idea: the reconstruction of the first "Iron Hand" of the Franconian imperial knight Götz von Berlichingen (1480–1562). We found that with this historical prosthesis, simple actions for daily use, such as holding a wine glass, a mobile phone, a bicycle handlebar grip, a horse’s reins, or some grapes, are possible without effort. Controlling this passive artificial hand, however, is based on the help of a healthy second hand.
The growing threat posed by multidrug-resistant (MDR) pathogens, such as Klebsiella pneumoniae (Kp), represents a significant challenge in modern medicine. Traditional antibiotic therapies are often ineffective against these pathogens, leading to high mortality rates. MDR Kp infections pose a novel challenge in military medical contexts, particularly in Medical Biodefense, as they can be deliberately spread, leading to resource-intensive care in military centres. Recognizing this issue, the European Defence Agency initiated a prioritised research project in 2023 (EDF Resilience PHAGE- SGA 2023). To address this challenge, the Bundeswehr Institute of Microbiology (IMB) leads BMBF- (Federal Ministry of Education and Research) and EU-funded projects on the use of bacteriophages as adjuvant therapy alongside antibiotics. Since 2017, the IMB has isolated and characterised Kp phages, collecting over 600 isolates and optimizing their production for therapy, in compliance with the EMA (European Medicine Agency) guidelines. This involves in vitro phage genome packaging to minimize endotoxin load, reduce manufacturing costs, and shorten production times. The goal of this work was to establish MinION sequencing (Oxford Nanopore Technology) as a quick and reliable way for initial identification and characterisation of phage genomes. Especially as a quick screening method for isolated on Kp, prior to more precise but also more expensive and time consuming sequencing methods like Illumina. This characterisation is crucial for developing a personalized pipeline aimed at producing magistral or Good Manufacturing Practice (GMP) quality medicinal phage solutions tailored individually for each patient. DNA extraction methods were compared to identify suitable input DNA for sequencing purposes. Additionally, the quality of this DNA was as- sessed to determine its suitability for in vitro phage packaging, which was successfully done achieving a phage titer of 103, confirming that the DNA used for MinION sequencing could indeed be used for acellular packaging. The created genomes were annotated and compared with Illumina sequencing, revealing high similarity in all five individually tested cases. Between the generated sequences only a 4% maximal percentual difference in genome size was observed, while simultaneously showing high similarity in the actual sequence. Throughout the course of this study, a total of 645.15 GB of sequencing data were generated. In total, 38 phages were successfully characterised, with 21 phage genomes assembled and annotated, and saved in the IMB database.
In 2015, Google engineer Alexander Mordvintsev presented DeepDream as technique to visualise the feature analysis capabilities of deep neural networks that have been trained on image classification tasks. For a brief moment, this technique enjoyed some popularity among scientists, artists, and the general public because of its capability to create seemingly hallucinatory synthetic images. But soon after, research moved on to generative models capable of producing more diverse and more realistic synthetic images. At the same time, the means of interaction with these models have shifted away from a direct manipulation of algorithmic properties towards a predominance of high level controls that obscure the model's internal working. In this paper, we present research that returns to DeepDream to assess its suit-ability as method for sound synthesis. We consider this research to be necessary for two reasons: it tackles a perceived lack of research on musical applications of DeepDream, and it addresses DeepDream's potential to combine data driven and algorithmic approaches. Our research includes a study of how the model architecture, choice of audio data-sets, and method of audio processing influence the acoustic characteristics of the synthesised sounds. We also look into the potential application of DeepDream in a live-performance setting. For this reason, the study limits itself to models consisting of small neural networks that process time-domain representations of audio. These models are resource-friendly enough to operate in real time. We hope that the results obtained so far highlight the attractiveness of Deep-Dream for musical approaches that combine algorithmic investigation with curiosity driven and open ended exploration.
This paper describes the authors' first experiments in creating an artificial dancer whose movements are generated through a combination of algorithmic and interactive techniques with machine learning. This approach is inspired by the time honoured practice of puppeteering. In puppeteering, an articulated but inanimate object seemingly comes to live through the combined effects of a human controlling select limbs of a puppet while the rest of the puppet's body moves according to gravity and mechanics. In the approach described here, the puppet is a machine-learning-based artificial character that has been trained on motion capture recordings of a human dancer. A single limb of this character is controlled either manually or algorithmically while the machine-learning system takes over the role of physics in controlling the remainder of the character's body. But rather than imitating physics, the machine-learning system generates body movements that are reminiscent of the particular style and technique of the dancer who was originally recorded for acquiring training data. More specifically, the machine-learning system operates by searching for body movements that are not only similar to the training material but that it also considers compatible with the externally controlled limb. As a result, the character playing the role of a puppet is no longer passively responding to the puppeteer but makes movement decisions on its own. This form of puppeteering establishes a form of dialogue between puppeteer and puppet in which both improvise together, and in which the puppet exhibits some of the creative idiosyncrasies of the original human dancer.
Generative machine learning models for creative purposes play an increasingly prominent role in the field of dance and technology. A particularly popular approach is the use of such models for generating synthetic motions. Such motions can either serve as source of ideation for choreographers or control an artificial dancer that acts as improvisation partner for human dancers. Several examples employ autoencoder-based deep-learning architectures that have been trained on motion capture recordings of human dancers. Synthetic motions are then generated by navigating the autoencoder's latent space. This paper proposes an alternative approach of using an autoencoder for creating synthetic motions. This approach controls the generation of synthetic motions on the level of the motion itself rather than its encoding. Two different methods are presented that follow this principle. Both methods are based on the interactive control of a single joint of an artificial dancer while the other joints remain under the control of the autoencoder. The first method combines the control of the orientation of a joint with iterative autoencoding. The second method combines the control of the target position of a joint with forward kinematics and the application of latent difference vectors. As illustrative example of an artistic application, this latter method is used for an artificial dancer that plays a digital instrument. The paper presents the implementation of these two methods and provides some preliminary results.
Steroid hormones (SHs) are a rising concern due to their high bioactivity, ubiquitous nature, and prolonged existence as a micropollutants in water, they pose a potential risk to both human health and the environment, even at low concentrations. Estrogens, progesterone, and testosterone are the three important types of steroids essential for human development and maintaining multiorgan balance, are focus to this concern. These steroid hormones originate
from various sources, including human and livestock excretions, veterinary medications, agricultural runoff, and pharmaceuticals, contributing to their presence in the environment. According to the recommendation of WHO, the guidance value for estradiol (E2) is 1 ng/L. There are several methods been attempted to remove the SH micropollutant by conventional water and wastewater technologies which are still under research. Among the various methods, electrochemical membrane reactor (EMR) is one of the emerging technologies that can address the challenge of insufficient SHs removal from the aquatic environment by conventional treatment. The degradation of SHs can be significantly influenced by various factors when treated with EMR.
In this project, the removal of SH and the important mechanism for the removal using carbon nanotube CNT-EMR is studied and the efficiency of CNT-EMR in treating the SH micropollutant is identified. By varying different parameters this experiment is carried out with the (PES-CNTs) ultrafiltration membrane. The study is carried out depending upon the SH removal based on the limiting factor such as cell voltage, flux, temperature, concentration, and type of the SH.
Batteries typically consist of multiple individual cells connected in series. Here we demonstrate single-cell state of charge (SOC) and state of health (SOH) diagnosis in a 24 V class lithium-ion battery. To this goal, we introduce and apply a novel, highly efficient algorithm based on a voltage-controlled model (VCM). The battery, consisting of eight single cells, is cycled over a duration of five months under a simple cycling protocol between 20 % and 100 % SOC. The cell-to-cell standard deviations obtained with the novel algorithm were 1.25 SOC-% and 1.07 SOH-% at beginning of cycling. A cell-averaged capacity loss of 9.9 % after five months cycling was observed. While the accuracy of single-cell SOC estimation was limited (probably owed to the flat voltage characteristics of the lithium iron phosphate, LFP, chemistry investigated here), single-cell SOH estimation showed a high accuracy (2.09 SOH-% mean absolute error compared to laboratory reference tests). Because the algorithm does not require observers, filters, or neural networks, it is computationally very efficient (three seconds analysis time for the complete data set consisting of eight cells with approx. 780.000 measurement points per cell).
This thesis focuses on the development and implementation of a Datagram Transport Layer Security (DTLS) communication framework within the ns-3 network simulator, specifically targeting the LoRaWAN model network. The primary aim is to analyse the behaviour and performance of DTLS protocols across different network conditions within a LoRaWAN context. The key aspects of this work include the following.
Utilization of ns-3: This thesis leverages ns-3’s capabilities as a powerful discrete event network simulator. This platform enables the emulation of diverse network environments, characterized by varying levels of latency, packet loss, and bandwidth constraints.
Emulation of Network Challenges: The framework specifically addresses unique challenges posed by certain network configurations, such as duty cycle limitations. These constraints, which limit the time allocated for data transmission by each device, are crucial in understanding the real-world performance of DTLS protocols.
Testing in Multi-client-server Scenarios: A significant feature of this framework is its ability to test DTLS performance in complex scenarios involving multiple clients and servers. This is vital for assessing the behaviour of a protocol under realistic network conditions.
Realistic Environment Simulation: By simulating challenging network conditions, such as congestion, limited bandwidth, and resource constraints, the framework provides a realistic environment for thorough evaluation. This allows for a comprehensive analysis of DTLS in terms of security, performance, and scalability.
Overall, this thesis contributes to a deeper understanding of DTLS protocols by providing a robust tool for their evaluation under various and challenging network conditions.
Strings P
(2021)
Strings is an audiovisual performance for an acoustic violin and two generative instruments, one for creating synthetic sounds and one for creating synthetic imagery. The three instruments are related to each other conceptually , technically, and aesthetically by sharing the same physical principle, that of a vibrating string. This submission continues the work the authors have previously published at xCoAx 2020. The current submission briefly summarizes the previous publication and then describes the changes that have been made to Strings. The P in the title emphasizes, that most of these changes have been informed by experiences collected during rehearsals (in German Proben). These changes have helped Strings to progress from a predominantly technical framework to a work that is ready for performance.
Strings
(2020)
This article presents the currently ongoing development of an audiovisual performance work with the title Strings. This work provides an improvisation setting for a violinist, two laptop performers, and two generative systems. At the core of Strings lies an approach that establishes a strong correlation among all participants by means of a shared physical principle. The physical principle is that of a vibrating string. The article discusses how this principle is used in both natural and simulated forms as main interaction layer between all performers and as natural or generative principle for creating audio and video.
Implementation and Evaluation of an Assisting Fuzzer Harness Generation Tool for AUTOSAR Code
(2024)
The digitalization in vehicles tends to add more connectivity such as over-the-air (OTA) updates. To achieve this digitization, each ECU (Electronic Control Unit) becomes smarter and needs to support more and more different externally available protocols such as TLS, which increases the attack surface for attackers. To ensure the security of a vehicle, fuzzing has proven to be an effective method to discover memory-related security vulnerabilities. Fuzzing the software run- ning on a ECU is not an easy task and requires a harness written by a human. The author needs a deep understanding of the specific service and protocol, which is time consuming. To reduce the time needed by a harness author, this thesis aims to develop FuzzAUTO, the first assistant harness generation tool targeting the AUTOSAR (AUTomotive Open System ARchitecture) BSW (Basic Software) to support manual harness generation.
The progress in machine learning has led to advanced deep neural networks. These networks are widely used in computer vision tasks and safety-critical applications. The automotive industry, in particular, has experienced a significant transformation with the integration of deep learning techniques and neural networks. This integration contributes to the realization of autonomous driving systems. Object detection is a crucial element in autonomous driving. It contributes to vehicular safety and operational efficiency. This technology allows vehicles to perceive and identify their surroundings. It detects objects like pedestrians, vehicles, road signs, and obstacles. Object detection has evolved from being a conceptual necessity to an integral part of advanced driver assistance systems (ADAS) and the foundation of autonomous driving technologies. These advancements enable vehicles to make real-time decisions based on their understanding of the environment, improving safety and driving experiences. However, the increasing reliance on deep neural networks for object detection and autonomous driving has brought attention to potential vulnerabilities within these systems. Recent research has highlighted the susceptibility of these systems to adversarial attacks. Adversarial attacks are well-designed inputs that exploit weaknesses in the deep learning models underlying object detection. Successful attacks can cause misclassifications and critical errors, posing a significant threat to the functionality and safety of autonomous vehicles. With the rapid development of object detection systems, the vulnerability to adversarial attacks has become a major concern. These attacks manipulate inputs to deceive the target system, significantly compromising the reliability and safety of autonomous vehicles. In this study, we focus on analyzing adversarial attacks on state-of-the-art object detection models. We create adversarial examples to test the models’ robustness. We also check if the attacks work on a different object detection model meant for similar tasks. Additionally, we extensively evaluate recent defense mechanisms to see how effective they are in protecting deep neural networks (DNNs) from adversarial attacks and provide a comprehensive overview of the most commonly used defense strategies against adversarial attacks, highlighting how they can be implemented practically in real-world situations.
Privacy is the capacity to keep some things private despite their social repercussions. It relates to a person’s capacity to control the amount, time, and circumstances under which they disclose sensitive personal information, such as a person’s physiology, psychology, or intelligence. In the age of data exploitation, privacy has become even more crucial. Our privacy is now more threatened than it was 20 years ago, outside of science and technology, due to the way data and technology highly used. Both the kinds and amounts of information about us and the methods for tracking and identifying us have grown a lot in recent years. It is a known security concern that human and machine systems face privacy threats. There are various disagreements over privacy and security; every person and group has a unique perspective on how the two are related. Even though 79% of the study’s results showed that legal or compliance issues were more important, 53% of the survey team thought that privacy and security were two separate things. Data security and privacy are interconnected, despite their distinctions. Data security and data privacy are linked with each other; both are necessary for the other to exist. Data may be physically kept anywhere, on our computers or in the cloud, but only humans have authority over it. Machine learning has been used to solve the problem for our easy solution. We are linked to our data. Protect against attackers by protecting data, which also protects privacy. Attackers commonly utilize both mechanical systems and social engineering techniques to enter a target network. The vulnerability of this form of attack rests not only in the technology but also in the human users, making it extremely difficult to fight against. The best option to secure privacy is to combine humans and machines in the form of a Human Firewall and a Machine Firewall. A cryptographic route like Tor is a superior choice for discouraging attackers from trying to access our system and protecting the privacy of our data There is a case study of privacy and security issues in this thesis. The problems and different kinds of attacks on people and machines will then be briefly talked about. We will explain how Human Firewalls and machine learning on the Tor network protect our privacy from attacks such as social engineering and attacks on mechanical systems. As a real-world test, we will use genomic data to try out a privacy attack called the Membership Inference Attack (MIA). We’ll show Machine Firewall as a way to protect ourselves, and then we’ll use Differential Privacy (DP), which has already been done. We applied the method of Lasso and convolutional neural networks (CNN), which are both popular machine learning models, as the target models. Our findings demonstrate a logarithmic link between the desired model accuracy and the privacy budget.