Authorship attribution has become increasingly accurate, posing a serious privacy risk for programmers who wish to remain anonymous. In this paper, we introduce SHIELD to examine the robustness of different code authorship attribution approaches against adversarial code examples. We define four attacks on attribution techniques, which include targeted and non-targeted attacks, and realize them using adversarial code perturbation. We experimented with a dataset of 200 programmers from the Google Code Jam competition to validate our methods. We target six state-of-the-art authorship attribution methods that adopt various techniques for extracting authorship traits from source code, including RNN, CNN, and code stylometry. Our experiments demonstrate the vulnerability of current authorship attribution methods against adversarial attacks. For the non-targeted attack, our experiments demonstrate the vulnerability of current authorship attribution methods against the attack with an attack success rate exceeding 98. 5% accompanied by a degradation of the identification confidence exceeding 13%. For targeted attacks, we show the possibility of impersonating a programmer using targeted adversarial perturbations with a success rate ranging from 66% to 88% for different authorship attribution techniques under several adversarial scenarios.

Federated Learning (FL) offers a privacy-preserving solution by enabling multiple clients to train a shared model collaboratively without centralizing data. However, the decentralized nature of FL presents challenges, particularly regarding security and performance under adversarial conditions. This paper investigates the effects of poisoning attacks under data heterogeneity. Our experiments evaluate the impact of varying malicious client fractions and poison concentration levels on the accuracy of the model. We explore the effects of poisoning attacks on FedAvg and FedNova models using medical imaging tasks. Our findings reveal that increasing data heterogeneity exacerbates the effects of poisoning, with FedNova demonstrating greater resilience compared to FedAvg. We found that the number of malicious clients plays a more significant role in degrading performance than the ratio of poisoning samples shared by each malicious client, suggesting that even modest levels of poisoning can be tolerated by most algorithms. The study highlights the importance of developing robust defense mechanisms to maintain model performance under adversarial conditions.

Social media has revolutionized communication, allowing people world-wide to connect and interact instantly. However, it has also led to increases in cyberbullying, which poses a significant threat to children and adolescents globally, affecting their mental health and well-being. It is critical to accurately detect the roles of individuals involved in cyberbullying incidents to effectively address the issue on a large scale. This study explores the use of machine learning models to detect the roles involved in cyberbullying interactions. After examining the AMiCA dataset and addressing class imbalance issues, we evaluate the performance of various models built with four underlying LLMs (i.e., BERT, RoBERTa, T5, and GPT-2) for role detection. Our analysis shows that oversampling techniques help improve model performance. The best model, a fine-tuned RoBERTa using oversampled data, achieved an overall F1 score of 83.5%, increasing to 89.3% after applying a prediction threshold. The top-2 F1 score without thresholding was 95.7%. Our method outperforms previously proposed models. After investigating the per-class model performance and confidence scores, we show that the models perform well in classes with more samples and less contextual confusion (e.g., Bystander Other), but struggle with classes with fewer samples (e.g., Bystander Assistant) and more contextual ambiguity (e.g., Harasser and Victim). This work highlights current strengths and limitations in the development of accurate models with limited data and complex scenarios.

Establishing trust and helping experts debug and understand the inner workings of deep learning models, interpretation methods are increasingly coupled with these models, building interpretable deep learning systems. However, adversarial attacks pose a significant threat to public trust by making interpretations of deep learning models confusing and difficult to understand. In this paper, we present a novel Single-class target-specific ADVersarial attack called SingleADV. The goal of SingleADV is to generate a universal perturbation that deceives the target model into confusing a specific category of objects with a target category while ensuring highly relevant and accurate interpretations. The universal perturbation is stochastically and iteratively optimized by minimizing the adversarial loss that is designed to consider both the classifier and interpreter costs in targeted and non-targeted categories. In this optimization framework, ruled by the first- and second-moment estimations, the desired loss surface promotes high confidence and interpretation scores of adversarial samples. By avoiding unintended misclassification of samples from other categories, SingleADV enables more effective targeted attacks on interpretable deep learning systems in both white-box and black-box scenarios. To evaluate the effectiveness of SingleADV, we conduct experiments using four different model architectures (ResNet-50, VGG-16, DenseNet-169, and Inception-V3) coupled with three interpretation models (CAM, Grad, and MASK). Through extensive empirical evaluation, we demonstrate that SingleADV effectively deceives target deep learning models and their associated interpreters under various conditions and settings. Our results show that the performance of SingleADV is effective, with an average attack success rate of 74% and prediction confidence exceeding 77% on successful adversarial samples. Furthermore, we discuss several countermeasures against SingleADV, including a transfer-based learning approach and existing preprocessing defenses.

With unfolding crises such as the COVID-19 pandemic, it is essential that factual information is dispersed at a rapid pace. One of the major setbacks to mitigating the effects of such crises is misinformation. Advancing technologies such as transformer-based architectures that can pick up underlying patterns and correlational information that constitutes information provide tools that can be used to identify what is misinformation/information. To identify and analyze the spread of misinformation, this work performs a quantitative analysis that uses X (previously Twitter) as the data source and a BERT-based model to identify misinformation. The information of the posts, users, and followers was collected based on hashtags and then processed and manually labeled. Furthermore, we tracked the spread of misinformation related to COVID-19 during the year 2021 and determined how communities that spread information and/or misinformation on social networks interact from an analytical perspective. Our findings suggest that users tend to post more misinformation than information, possibly intentionally spreading misinformation. Our model showed good performance in classifying tweets as information/misinformation, resulting in an accuracy of 86%.

Social media continues to have an impact on the trajectory of humanity. However, its introduction has also weaponized keyboards, allowing the abusive language normally reserved for in-person bullying to jump onto the screen, i.e., cyberbullying. Cyberbullying poses a significant threat to adolescents globally, affecting the mental health and well-being of many. A group that is particularly at risk is the LGBTQ+ community, as researchers have uncovered a strong correlation between identifying as LGBTQ+ and suffering from greater online harassment. Therefore, it is critical to develop machine learning models that can accurately discern cyberbullying incidents as they happen to LGBTQ+ members. The aim of this study is to compare the efficacy of several transformer models in identifying cyberbullying targeting LGBTQ+ individuals. We seek to determine the relative merits and demerits of these existing methods in addressing complex and subtle kinds of cyberbullying by assessing their effectiveness with real social media data.

Traditional one-time authentication mechanisms cannot authenticate smartphone users’ identities throughout the session – the concept of using behavioral-based biometrics captured by the built-in motion sensors and touch data is a candidate to solve this issue. Many studies proposed solutions for behavioral-based continuous authentication; however, they are still far from practicality and generality for real-world usage. To date, no commercially deployed implicit user authentication scheme exists because most of those solutions were designed to improve detection accuracy without addressing real-world deployment requirements. To bridge this gap, we tackle the limitations of existing schemes and reach toward developing a more practical implicit authentication scheme, dubbed MotionID, based on a one-class detector using behavioral data from motion sensors when users touch their smartphones. Compared with previous studies, our work addresses the following challenges: 1) Global mobile average to dynamically adjust the sampling rate for sensors on any device and mitigate the impact of using sensors’ fixed sampling rate; 2) Over-all-apps to authenticate a user across all the mobile applications, not only on-specific application; 3) Single-device-evaluation to measure the performance with multiple users’ (i.e., genuine users and imposters) data collected from the same device; 4) Rapid authentication to quickly identify users’ identities using a few samples collected within short durations of touching (1–5 s) the device; 5) Unconditional settings to collect sensor data from real-world smartphone usage rather than a laboratory study. To show the feasibility of MotionID for those challenges, we evaluated the performance of MotionID with ten users’ motion sensor data on five different smartphones under various settings. Our results show the impracticality of using a fixed sampling rate across devices that most previous studies have adopted. MotionID is able to authenticate users with an F1-score up to 98.5% for some devices under practical requirements and an F1-score up to roughly 90% when considering the drift concept and rapid authentication settings. Finally, we investigate time efficiency, power consumption, and memory usage considerations to examine the practicality of MotionID.

Deep Learning (DL) is rapidly maturing to the point that it can be used in safety- and security-crucial applications, such as self-driving vehicles, surveillance, drones, and robots. However, adversarial samples, which are undetectable to the human eye, pose a serious threat that can cause the model to misbehave and compromise the performance of such applications. Addressing the robustness of DL models has become crucial to understanding and defending against adversarial attacks. In this study, we perform a series of experiments to examine the effect of adversarial attacks and defenses on various model architectures across well-known datasets. Our research focuses on black-box attacks such as SimBA, HopSkipJump, MGAAttack, and boundary attacks, as well as preprocessor-based defensive mechanisms, including bits squeezing, median smoothing, and JPEG filter. Experimenting with various models, our results demonstrate that the level of noise needed for the attack increases as the number of layers increases. Moreover, the attack success rate decreases as the number of layers increases. This indicates that model complexity and robustness have a significant relationship. Investigating the diversity and robustness relationship, our experiments with diverse models show that having a large number of parameters does not imply higher robustness. Our experiments extend to show the effects of the training dataset on model robustness. Using various datasets such as ImageNet-1000, CIFAR-100, and CIFAR-10 are used to evaluate the black-box attacks. Considering the multiple dimensions of our analysis, e.g., model complexity and training dataset, we examined the behavior of black-box attacks when models apply defenses. Our results show that applying defense strategies can significantly reduce attack effectiveness. This research provides in-depth analysis and insight into the robustness of DL models against various attacks, and defenses.

Rapid advancements of deep learning are accelerating adoption in a wide variety of applications, including safety-critical applications such as self-driving vehicles, drones, robots, and surveillance systems. These advancements include applying variations of sophisticated techniques that improve the performance of models. However, such models are not immune to adversarial manipulations, which can cause the system to misbehave and remain unnoticed by experts. The frequency of modifications to existing deep learning models necessitates thorough analysis to determine the impact on models' robustness. In this work, we present an experimental evaluation of the effects of model modifications on deep learning model robustness using adversarial attacks. Our methodology involves examining the robustness of variations of models against various adversarial attacks. By conducting our experiments, we aim to shed light on the critical issue of maintaining the reliability and safety of deep learning models in safety- and security-critical applications. Our results indicate the pressing demand for an in-depth assessment of the effects of model changes on the robustness of models.

Deep learning methods have gained increasing attention in various applications due to their outstanding performance. For exploring how this high performance relates to the proper use of data artifacts and the accurate problem formulation of a given task, interpretation models have become a crucial component in developing deep learning-based systems. Interpretation models enable the understanding of the inner workings of deep learning models and offer a sense of security in detecting the misuse of artifacts in the input data. Similar to prediction models, interpretation models are also susceptible to adversarial inputs. This work introduces two attacks, AdvEdg and AdvEdge+, which deceive both the target deep learning model and the coupled interpretation model. We assess the effectiveness of proposed attacks against four deep learning model architectures coupled with four interpretation models that represent different categories of interpretation models. Our experiments include the implementation of attacks using various attack frameworks. We also explore the attack resilience against three general defense mechanisms and potential countermeasures. Our analysis shows the effectiveness of our attacks in terms of deceiving the deep learning models and their interpreters, and highlights insights to improve and circumvent the attacks.

Both CT and MRI images are acquired for HDR prostate brachytherapy patients at our institution. CT is used to identify catheters and MRI is used to segment the prostate. To address scenarios of limited MRI access, we developed a novel Generative Adversarial Network (GAN) to generate synthetic MRI (sMRI) from CT with sufficient soft-tissue contrast to provide accurate prostate segmentation without MRI (rMRI).

Target and organ delineation during prostate high-dose-rate (HDR) brachytherapy treatment planning can be improved by acquiring both a postimplant CT and MRI. However, this leads to a longer treatment delivery workflow and may introduce uncertainties due to anatomical motion between scans. We investigated the dosimetric and workflow impact of MRI synthesized from CT for prostate HDR brachytherapy. Seventy-eight CT and T2-weighted MRI datasets from patients treated with prostate HDR brachytherapy at our institution were retrospectively collected to train and validate our deep-learning-based image-synthesis method. Synthetic MRI was assessed against real MRI using the dice similarity coefficient (DSC) between prostate contours drawn using both image sets. The DSC between the same observer's synthetic and real MRI prostate contours was compared with the DSC between two different observers’ real MRI prostate contours. New treatment plans were generated targeting the synthetic MRI-defined prostate and compared with the clinically delivered plans using target coverage and dose to critical organs. Variability between the same observer's prostate contours from synthetic and real MRI was not significantly different from the variability between different observer's prostate contours on real MRI. Synthetic MRI-planned target coverage was not significantly different from that of the clinically delivered plans. There were no increases above organ institutional dose constraints in the synthetic MRI plans.

With increased privacy concerns, anonymity tools such as VPNs and Tor have become popular. However, the packet metadata such as the packet size and number of packets can still be observed by an adversary. This is commonly known as fingerprinting and website fingerprinting attacks have received a lot of attention recently as a known victim's website visits can be accurately predicted, deanonymizing that victim's web usage. Most of the previous work have been performed in laboratory settings and have made two assumptions: 1) a victim visits one website at a time, and 2) the whole website visit with all the network packets can be observed. To validate these assumptions, a new private webbrowser extension called WebTracker is deployed with real users. WebTracker records the websites visited, when the website loading starts, and when the website loading finishes. Results show that users' browsing patterns are different than what was previously assumed. Users may browse the web in a way that acts as a countermeasure against website fingerprinting due to multiple websites overlapping and downloading at the same time. Over 15% of websites overlap with at least one other website and each overlap was 66 seconds. Moreover, each overlap happens roughly 9 seconds after the first website download has started. Thus, this reinforces some previous work that the beginning of a website is more important than the end for a website fingerprinting attack.

With the DNS traffic, an eavesdropper can easily identify websites that a user is visiting. In order to address this concern of web privacy, encryption is used by performing a DNS lookup over HTTPS. In this paper, we studied whether the encrypted DoH traffic could be exploited to identify websites that a user has visited. This is a different type of website fingerprinting by analyzing encrypted DNS network traffic rather than the network traffic between the client and the web server. DNS typically uses fewer network packets than a website download. Our model and algorithm can accurately predict one out of 10,000 websites with a 95% accuracy using the first 50 DoH packets. In the open-world environment with 100,000 websites, our model achieves an F1-score of 93%.

This paper analyzes the reaction of social network users in terms of different aspects including sentiment analysis, topic detection, emotions, and the geo-temporal characteristics of our dataset. We show that the dominant sentiment reactions on social media are neutral, while the most discussed topics by social network users are about health issues. This paper examines the countries that attracted a higher number of posts and reactions from people, as well as the distribution of health-related topics discussed in the most mentioned countries. We shed light on the temporal shift of topics over countries. Our results show that posts from the top-mentioned countries influence and attract more reactions worldwide than posts from other parts of the world.

We have developed a PixCycleGAN framework to generate sMRI from CT with enhanced soft-tissue contrast on the boundary of prostate for HDR treatment planning. For our study, the DSC and HD of the prostate contours on sMRI vs. rMRI agrees within IOV. In the future, we will evaluate dosimetric impact of using sMRI in prostate segmentation for HDR treatment planning.

This work proposed W-Net, a CNN-based architecture with a small number of layers, to accurately classify the five WBC types. We evaluated W-Net on a real-world large-scale dataset and addressed several challenges such as the transfer learning property and the class imbalance. W-Net achieved an average classification accuracy of 97%. We synthesized a dataset of new WBC image samples using DCGAN, which we released to the public for education and research purposes.

This research utilizes various characterizing features extracted from each malware using static and dynamic analysis to build seven machine learning models to detect and analyze packed windows malware. We use a large-scale dataset of over 107,000 samples covering unpacked and packed malware using ten different packers. We examined the performance of seven machine learning techniques using 50 dynamic and static features. Our results show that packed malware can circumvent detection when a single analysis is performed while applying both static and dynamic methods can help improve the detection accuracy around 2% to 3%.

Deep neural network models are susceptible to malicious manipulations even in the black-box settings. Providing explanations for DNN models offers a sense of security by human involvement, which reveals whether the sample is benign or adversarial even though previous studies achieved a high attack success rate. However, interpretable deep learning systems (IDLSes) are shown to be susceptible to adversarial manipulations in white-box settings. Attacking IDLSes in black-box settings is challenging and remains an open research domain. In this work, we propose a black-box version of the white-box AdvEdge approach against IDLSes, which is query-efficient and gradient-free without obtaining any knowledge of the target DNN model and its coupled interpreter. Our approach takes advantage of transfer-based and score-based techniques using the effective microbial genetic algorithm (MGA). We achieve a high attack success rate with a small number of queries and high similarity in interpretations between adversarial and benign samples.

The rapid pace of malware development and the widespread use of code obfuscation, polymorphism, and morphing techniques pose a considerable challenge to detecting and analyzing malware. Today, it is difficult for antivirus applications to use traditional signature-based detection methods to detect morphing malware. Thus, the emergence of structure graph-based detection methods has become a hope to solve this challenge. In this work, we propose a method for detecting malware using graphs’ spectral heat and wave signatures, which are efficient and size- and permutation-invariant. We extracted 250 and 1,000 heat and wave representations, and we trained and tested heat and wave representations on eight machine learning classifiers. We used a dataset of 37,537 unpacked Windows malware executables and extracted the control flow graph (CFG) of each Windows malware to obtain the spectral representations. Our experimental results showed that by using heat and wave spectral graph theory, the best malware analysis accuracy reached 95.9%.

There is an increasing demand for comprehensive and in-depth analysis of behaviors of various attacks and the possible defenses against common deep learning models under several adversarial scenarios. In this study, we conducted four separate investigations. First, we examine the relationship between the model’s complexity and its robustness against the studied attacks. Second, the connection between the performance and diversity of models is examined. Third, the first and second experiments were tested across different datasets to explore the impact of the dataset on the performance of the model. Four, throughout the defense strategies, the model behavior is extensively investigated.

In this article, we survey more than 140 recent behavioral biometric-based approaches for continuous user authentication, including motion-based methods (28 studies), gait-based methods (19 studies), keystroke dynamics-based methods (20 studies), touch gesture-based methods (29 studies), voice-based methods (16 studies), and multimodal-based methods (34 studies). The survey provides an overview of the current state-of-the-art approaches for continuous user authentication using behavioral biometrics captured by smartphones’ embedded sensors, including insights and open challenges for adoption, usability, and performance.