Dark patterns, i.e., deceptive design patterns that employ manipulative strategies to deceive or force online users to take a decision against their interests, are widely available in online world. However, the variety and quantity of structured and labelled dark pattern datasets, which are critical for automated dark pattern detection, particularly for AI-based detection models, are limited. In this study, we leverage Large Language Models’ (LLM) sophisticated text data generation ability and propose a dark pattern text data augmentation method by utilizing a state of art open source language model and multi-agents framework, which has generator and controller models. Evaluation of the augmentation demonstrates that while increasing the data size, our proposal-based augmented data preserves the same dark pattern characteristics of the source data and maintains its diversity. We set forth that dark pattern text data can be generated even based on a few examples via prompt engineering techniques on the LLMs. We also show that our augmented data can be used to fine-tune pre-trained language models using Low-Rank Adaptation to enhance their robustness in detecting dark patterns.

To detect deceptive design patterns on UIs, traditional artificial intelligence models, such as machine learning, have limited coverage and a lack of multimodality. In contrast, the capabilities of Multimodal Large Language Model (MM-LLM) can achieve wider coverage with superior performance in the detection, while providing reasoning behind each decision. We propose and implement an MM-LLM-based approach (DeceptiLens) that analyzes UIs and assesses the presence of deceptive design patterns. We utilize Retrieval Augmented Generation (RAG) process in our design and task the model with capturing the deceptive patterns, classifying its category, e.g., false hierarchy, confirmshaming, etc., and explaining the reasoning behind the classifications by employing recent prompt engineering techniques, such as Chain-of-Thought (CoT). We first create a dataset by collecting UI screenshots from the literature and web sources and quantify the agreement between the model's outputs and a few experts' opinions. We additionally ask experts to gauge the transparency of the system's explanations for its classifications in terms of recognized metrics of clarity, correctness, completeness, and verifiability. The results indicate that our approach is capable of capturing the deceptive patterns in UIs with high accuracy while providing clear, correct, complete, and verifiable justifications for its decisions. We additionally release two curated datasets, one with expert-labeled UIs with deceptive design patterns, and one with AI-based generated explanations. Lastly, we propose recommendations for future improvement of the approach in various contexts of use.

In recent years, the importance of computer security has increased due to the rapid advancement of digital technology, widespread Internet use, and increased sophistication of cyberattacks. Machine learning has gained great interest in securing data systems because it offers the capability of automatically detecting and responding to security threats in real time, which is crucial for maintaining the security of computer systems and protecting data from malicious attacks. This study concentrates on phishing attack detection systems, a prevalent cyber-threat. These systems assess the features of the incoming requests to identify whether they are malicious or not. Although the number of features is increasing in these systems, feature selection has become an essential pre-processing phase that identifies the most important features of a set of available features to prevent overfitting problems, improve model performance, reduce computational cost, and decrease training and execution time. Leveraging genetic algorithms, known for simulating natural selection to identify optimal solutions, we propose a novel feature selection method, based on genetic algorithms and locally optimized, that is applied to a URL-based phishing detection system with machine learning models. Our research demonstrates that the proposed technique offers a promising strategy for improving the performance of machine learning models.

We propose an architectural design for a Trust Level Evaluation Engine. The engine is meant to work in a complex and dynamic environment of potentially untrustworthy sources of information where the situational knowledge is partial and subjective from the viewpoint of the information source, thus potentially inconsistent and contradictory. Consistently with a Zero-Trust approach, no initial trust between nodes should be assumed, since a decision-making module shall nevertheless figure out its level of confidence about the truth of a proposition over the reality. Our design is theory-agnostic and can be instantiated on different mathematical subjective model theories, but we demonstrate its feasibility by mapping it on the Subjective Logic. We also discuss critical design choices and algorithmic details that are only partially addressed in the abstract description of the theory, and we demonstrate how the engine effectively works on large and complex subjective trust networks. Additionally, we offer a proof-of-concept implementation to showcase the proposed architecture’s ability to handle intricate and complex networks.

Dark patterns are deceptive design elements of digital choice architectures that are implemented to drive users’ actions towards decisions that are not necessarily in their best interest, such as accepting privacy-invasive practices. Most dark patterns are considered unlawful, but their description is rather informal. Thus, detecting dark patterns among the various existing design patterns and discerning what is an illegitimate design practice may depend on the subjective interpretation of expert users (such as regulators, civil society organizations, and academic researchers) who may not fully agree. The need to ground any evaluation on evidence calls for a reliable approach that is based on descriptions relying on observable, measurable features. Taking cookie consent as a use case, where dark patterns are ubiquitous and intensively under scrutiny, we propose a systematic approach to describe the characteristics of deceptive design patterns that are intended to reconcile the interpretations of expert users. In particular: i) we identify use case-specific dark pattern types using the ontology drafted by Gray et al. (2024); ii) we clarify the relationships between those types and the dark patterns’ attributes proposed by Mathur et al. (2021); iii) we propose a list of observable and measurable user-interaction features of dark patterns covering visual, process, and language design aspects, iv) we describe the attributes based on our measurable features to lower the subjectivity of users’ interpretation. Finally, we discuss our proposal’s cross-domain applicability and the potential for future work, including how to improve the descriptions of the attributes via semiformal languages, to generate an objective and usable framework to assess the presence of deceptive design patterns in digital interfaces.

There has been a burst of discussions about how to characterize and recognize online dark patterns — i.e., web design strategies that aim to steer user choices towards what favours service providers or third parties like advertisers rather than what is in the best interest of users. Dark patterns are common in cookie banners where they are used to influence users to accept being tracked for more purposes than a data protection by default principle would dictate. Despite all the discussions, an objective, transparent, and verifiable assessment of dark patterns’ qualities is still missing. We contribute to bridging this gap by studying several cookie processes, in particular their multi-layered information flow —that we represent as message sequence charts—, and by identifying a list of observable and measurable features that we believe can help describing the presence of dark patterns in digital consent flows. We propose thirty one of such properties that can be operationalised into metrics and therefore into objective procedures for the detection of dark patterns.

This article discusses the challenges of pseudonymizing unstructured, noisy social media data for cybersecurity research purposes and presents an opensource package developed to pseudonymize personal and confidential information (i.e., personal names, companies, and locations) contained in such data. Its goal is to facilitate compliance with EU data protection obligations and the upholding of research ethics principles like the respect for the autonomy, privacy and dignity of research participants, the social responsibility of researchers, and scientific integrity. We discuss the limitations of the pseudonymizer package, their ethical import, and the additional security measures that should be adopted to protect the confidentiality of the data.

Phishing attacks are one of the most preferred types of attacks for cybercriminals, who can easily contact a large number of victims through the use of social networks, particularly through email messages. To protect end users, most of the security mechanisms control Uniform Resource Locator (URL) addresses because of their simplicity of implementation and execution speed. However, due to sophisticated attackers, this mechanism can miss some phishing attacks and has a relatively high false positive rate. In this research, a hybrid technique is proposed that uses not only URL features, but also content-based features as the second level of detection mechanism, thus improving the accuracy of the detection system while also minimizing the number of false positives. Additionally, most phishing detection algorithms use datasets that contain easily differentiated data pieces, either phishing or legitimate. However, in order to implement a more secure protection mechanism, we aimed to collect a larger and high-risk dataset. The proposed approaches were tested on this High-Risk URL and Content-Based Phishing Detection Dataset that only contains suspicious websites from PhishTank. According to experimental studies, an accuracy rate of 98.37 percent was achieved on a more realistic dataset for phishing detection.

1 Abstract —Phishing attacks are one of the most preferred types of attacks for cybercriminals, who can easily contact a large number of victims through the use of social networks, particularly through email messages. To protect end users, most of the security mechanisms control Uniform Resource Locator (URL) addresses because of their simplicity of implementation and execution speed. However, due to sophisticated attackers, this mechanism can miss some phishing attacks and has a relatively high false positive rate. In this research, a hybrid technique is proposed that uses not only URL features, but also content-based features as the second level of detection mechanism, thus improving the accuracy of the detection system while also minimizing the number of false positives. Additionally, most phishing detection algorithms use datasets that contain easily differentiated data pieces, either phishing or legitimate . However, in order to implement a more secure protection mechanism, we aimed to collect a larger and high-risk dataset. The proposed approaches were tested on this High-Risk URL and Content-Based Phishing Detection Dataset that only contains suspicious websites from PhishTank. According to experimental studies, an accuracy rate of 98.37 percent was achieved on a more realistic dataset for phishing detection.

The Internet is getting stronger day by day and it makes our lives easier with many applications that are executed on cyberworld. However, with the development of the internet, cyber-attacks have increased gradually and identity thefts have emerged. It is a type of fraud committed by intruders by using fake web pages to access people's private information such as userid, password, credit card number and bank account numbers, etc. These scammers can also send e-mail from many important institutions and organizations by using phishing attacks which imitate these web pages and acts as if they are original. Traditional security mechanisms can not prevent these attacks because they directly target the weakest part of connection: end-users. Machine learning technology has been used to detect and prevent this type of intrusions. The anti-phishing method has been developed by detecting the attacks made with the technologies used. In this paper, we combined the websites used by phishing attacks into a dataset, then we obtained some results using 4 classification algorithms with this dataset. The experimental results showed that the proposed systems give very good accuracy levels for the detection of these attacks.

Due to the growing trend of Internetization, the number of connected computers has been increasing day by day. Almost all companies are transferring their main operations from the real world to the cyberworld. Although this increases the marketplace of the firms, it also brings lots of vulnerabilities, such as cyber-attacks, especially with the anonymous structure of the Internet. Phishing is one of the popular attack types which exploits the vulnerabilities to user unawareness. There are some works in the literature that gets help from the rule-based detection systems as a static preventions mechanism, and machine learning-based systems as dynamic prevention mechanisms. In this work, we implemented a deep neural network (DNN) based phishing detection system by analyzing the URL of the suspicious websites. Although in almost all previous researches the used datasets are collected by different resources in which legitimate and phishing websites are clear, in this research, we firstly create a high-risk dataset, which contains only the suspicious websites which are reported to PhishTank website. Experimental research showed that the proposed system gives a very good level of efficiency both in accuracy and execution time manner.

Recently, cyber-attacks have increased worldwide, especially during the pandemic period. The number of connected devices in the world and the anonymous structure of the internet enable this security deficit for not only computer networks but also single computing devices. With the connected use of computing device in anytime and anywhere conditions, lots of real-world activities are transferred to the digital world by adapting them to new lifestyles. Thus, the concept of cybersecurity has become more focused not only for security admins but also for academicians/researchers. Phishing attacks, which hackers mostly prefer to use in the last decade, have become even more harmful because its focuses on the weakest part of the security chain: computer user. Therefore, it is extremely important to prevent these cyber-attacks before they reach users. Based on this idea, we aimed to implement a phishing detection system by using a Convolutional Neural Network with n-gram features that are extracted from URLs. There are different n-gram feature extraction techniques, and in this work, it is aimed to determine which of them is more effective for our proposals. As a second goal, it is aimed to discover what parameters of the n-gram work best. In experiments, it is discovered that unigram has the highest accuracy rate. It was observed that, instead of all the characters that are obtained in unigram, the specified 70 characters (regardless of case sensitivity) give the highest accuracy rate of 88.90% with a High-Risk URL dataset. Experimental results also showed that a URL can be classified (either as legitimate or phishing) in about 0.008 seconds. These metrics can be accepted at a very good rate both in accuracy and run-time efficiency.

Both the population growth in the world and the number of people living in the city are increasing day by day while the rural population decreases. However, lack of resources, limited sources or difficulties in city life also arise. In response to these situations, the Smart City concept emerges and thanks to this concept, which integrates information and communication technology, it optimizes the available resources and makes people's life more qualified, efficient and comfortable. Smart cities include the phases of transferring, storing and processing real-time data from sensors and manage variable subsystems. While the transactions are carried out successfully for the purposes, the security and privacy problems must be examined carefully. In this paper, we analyze security issues and challenges in two ways: IoT-based and cloud-based. And as a solution proposal, it is stated that we can use authentication, encrypted communication, and blockchain to ensure data privacy. It is difficult to provide security in smart city architectures due to its decentralized and distributed structure with the existing technology IoT. Blockchain offers a security approach with a decentralized and distributed structure.

According to the United Nations' expectation, the total population of the cities will be doubled in the next three decades. This accelerating growth causes crucial problems in the main components of both traditional cities and smart cities. To increase the living quality of the residence in smart cities, enabling a clean, healthy, and sustainable environment are the major fields for the smart cities' managers and directors. One of the main infrastructures of the smart city is identified as smart health, which can be enabled with the use of modern technologies such as Internet of Things, especially for accessing the patients when they need help. In this Project, a smart ambulance management system is proposed in a smart city environment. If a patient needs an ambulance, the operator finds the nearest ambulance and direct it to the patient. The coordinates of ambulances are dynamically traced by the system, and Google Maps, as a third-party service, is used in order to calculate the shortest path to the casualty. After reaching to the patient, the expert (doctor or nurse) investigates the situation and finds the best available hospital by the proposed system. The experimental results showed that the proposed system finds the best solution in an acceptable $time$.

Considerable growth of computer science has created novel solutions for variable problem fields and has increased the efficiency of available solutions. Evolutionary algorithms are quite successful in dealing with real-world problems that require optimization. In this article, we implemented a Genetic Algorithm that is well known evolutionary algorithm in order to provide an efficient solution for the Distribution of Pharmaceutical Products, which is a vital optimization problem, especially in situations such as a pandemic. The Multiple Traveling Salesman Problem approach was used to distribute pharmaceutical products as soon as possible. Moreover, we strengthened our proposal algorithm with 2-Opt Algorithm to get optimal results in earlier iterations. Different datasets from a library were applied to measure the quality of solutions and computation time. At the end of the work, we observed that our proposed algorithm generates successful solutions in an acceptable running time. This study will be extended with a new mutation concept as future work.

Due to the digitalization of our lives and the connection of almost all types of devices to the Internet, cybersecurity has become one of the hottest research areas in computer engineering. The anonymous structure of the Internet results in inevitable increase not only in the number of attacks but also in their types. Attackers generally prefer the use of webpages for deceiving the end-users. These pages are designed as if they are legitimate webpages, and they aimed to reach some important credentials of their victims, such as userIds, passwords, credit card numbers, etc. Therefore, the systems which detect and decrease this type of cyber threat can easily reduce the number of victims/losses and also increase the security of the computer network. To catch this type of malicious web pages, different approaches are proposed in the literature. Due to the dynamic structure of Machine Learning (ML) approaches, they are generally preferred for the detection of these pages. In this paper, we proposed a cyber threat detection system, which analyzes the content of the web pages to understand whether it is a legitimate or malicious page. The experimental results depict that the proposed model detects the malicious web pages with a very good level of accuracy.

With the growth of population, there is an inevitable increase in solid waste especially in urban areas. For the municipalities, especially in smart cities, this becomes a major problem in nature, and it leads to many socio-economic and environmental problems. Thus, lowering our living standards. To eliminate or minimize these problems, using the Internet of Things (IOT) technology is the most advantageous solution for collecting solid wastes within this scope. In this paper, we proposed an optimal waste collection mechanism with the use of some IoT devices in the garbage cans which show the level of waste in them. For testing the proposal, we select a sample environment as a specific region of Istanbul, which is named as Bakirkoy. With the use of sensors, it is aimed to detect which cans are needed to be visited. Then with the use of an evolutionary algorithm, Genetic Algorithm, best path for visiting these cans can be planned in a very short time. By using this approach, it is aimed to effectively use the workforce/resources of the smart cities and making less traffic jam on the roads. Experimental results showed that the proposed system results very good enhancement in the waste collection operations.

Unmanned Systems has been taking place of manned systems in several fields like aviation. Unmanned Aerial Vehicle (UAV), one of the most popular and effective unmanned systems, is gradually becoming the vital element of aviation because of its high success rate in both military and civilian missions. Basic problem of UAV is finding the best path in tough environment. Coverage zones of radars and complex environment are the main obstacles in this problem. A UAV intends to travel all control points in an optimal way to be more productive while avoiding radars. In this paper, we used Genetic Algorithm (GA), which is Evolutionary algorithm, to find the optimal flyable path for the UAVs in a 3D environment. Each generation is anticipated to be better than its previous generation in GA. For the purpose of reaching an optimal path, solving the Travelling Salesman Problem (TSP) is one of the major phases in the proposed method. In order to show the visual of solution in better quality, we preferred MATLAB as the implementation environment. Additionally, there is a shared library and mathematical calculations are easier in MATLAB. The complexity of our problem can be increased by adding extra constraints caused by the dynamic environment as the future works. Experimental results show that GA can be opted for optimal path planning for the UAVs.