Franziska Schwarz

Mrs. Franziska Schwarz

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 February 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
FSFranziska Schwarz
Franziska Schwarz
KSKlaus Schwarz
Klaus Schwarz
Daniel Arias ArandaDaniel Arias Aranda
Daniel Arias Aranda
KBKendrick Bollens
Kendrick Bollens
NSNavaneeth Shivananjappa
Navaneeth Shivananjappa
VDVesna Dimitrova
Vesna Dimitrova
Rapid classification of social media content during humanitarian crises is essential for effective disaster relief; however, traditional approaches require extensive annotated training data, which are often unavailable during new disasters. This paper presents a training-free, multimodal classification framework that leverages zero-shot vision-language models to analyze disaster-related social media content without task-specific training. The framework employs a two-stage prompt-engineered pipeline using the locally deployable Mistral-Small-3.1-24B-Instruct model, performing binary informativeness detection followed by multiclass categorization into eight humanitarian categories through structured JSON output generation. Evaluation on the CrisisMMD dataset of 18,082 multimodal samples from seven natural disasters demonstrated binary F1 scores above 0.84 for both text and image informativeness detection and weighted F1 scores of 0.61 (text) and 0.72 (image) for humanitarian categorization. The framework generalizes consistently across all disaster types with minimal performance variance (standard deviation below 0.031) and operates entirely on local infrastructure without cloud dependencies, requiring only moderate GPU resources. By eliminating training data requirements, this approach enables immediate deployment during new disasters, demonstrating that zero-shot multimodal classification achieves practically relevant performance for real-time crisis response.

21 January 2024 in Electronic Imaging
Verified
Journal Article
Open access
PDF file
SMSamrudha Mhatre
Samrudha Mhatre
FSFranziska Schwarz
Franziska Schwarz
KSKlaus Schwarz
Klaus Schwarz
RCReiner Creutzburg
Reiner Creutzburg
Open-source technologies (OSINT) and Social Media Intelligence (SOCMINT) are becoming increasingly popular with investigative and government agencies, intelligence services, media companies, and corporations - but also for cybercriminals in email phishing. The amount of public and private data available is rising rapidly. OSINT and SOCMINT technologies use sophisticated techniques and special tools to analyze the continually growing sources of information efficiently. This work aims to find descriptive information using the OSINT tools available online. The target will be achieved with the help of dummy accounts that would help understand the tools and evaluate further different tools. Also, find out what tools are commonly used and what improvements can be made to make them more descriptive for analysts.

21 January 2024 in Electronic Imaging
Verified
Journal Article
Open access
PDF file
KSKlaus Schwarz
Klaus Schwarz
FSFranziska Schwarz
Franziska Schwarz
KBKnud Brandis
Knud Brandis
RCReiner Creutzburg
Reiner Creutzburg
This paper introduces AI-based Cybersecurity Management Consulting (AI-CMC) as a disruptive technology to address the growing complexity of cybersecurity threats. AI-CMC combines advanced AI techniques with cybersecurity management, offering proactive and adaptive strategies through machine learning, natural language processing, and big data analytics. It enables real-time threat detection, predictive analytics, and intelligent decision-making. The paper explores AI-CMC's data-driven approach, learning models, and collaborative framework, demonstrating its potential to revolutionize cyber-security. It examines AI-CMC's benefits, challenges, and ethical considerations, emphasizing transparency and bias mitigation. A roadmap for transitioning to AI-CMC and its implications for industry standards, policies, and global strategies are discussed. Despite potential limitations and vulnerabilities, AI-CMC offers transformative solutions for enhancing threat resilience and safeguarding digital assets, calling for collaborative efforts and responsible use for a secure digital future.

16 January 2022 in Electronic Imaging
Verified
Journal Article
Open access
PDF file
FSFranziska Schwarz
Franziska Schwarz
KSKlaus Schwarz
Klaus Schwarz
RCReiner Creutzburg
Reiner Creutzburg
Morphing is a well-researched topic in computer graphics and image processing. Unlike cross-fading, morphing transforms a source image into a target image using distortions and the adjustment of predefined features (control points). This transition is intended to create results that are as realistic as possible, and it happens notably when the face on the source and target images does not differ too much. Therefore, the typical morphing process consists of warping essential image elements (e.g., facial features such as eyes, mouth, and facial contours) in the source and target image with the help of selected control points in such a way that these areas can be brought into conformity with each other or align themselves. For effects as close to reality as possible, the source and target images must not differ too much. The training course developed for this work is intended to introduce the dangers posed by the malicious manipulation of facial photographs, the so-called morphing attack. An algorithm matches the images of two different people so that the resulting facial image combines the identification features of both people. Research has shown that these images are difficult for the human eye to distinguish from real, unaltered photographs. The face morphing attack exploits the weakness in the application for identification. If a morphed passport photo remains undiscovered, a genuine identity document with a manipulated photo is issued that may allow two different persons to cross a border without authorization. This targeted training course consists of 10 modules for a one-week training course and is designed to help identify morphed facial images and reduce their acceptance at ID and passport control.