This paper presents a keystroke-based framework for detecting LLM-assisted cheating in Korean, addressing key gaps in prior research regarding language coverage, cognitive context, and the granularity of LLM involvement. Our proposed dataset includes 69 participants who completed writing tasks under three conditions: Bona fide writing, paraphrasing ChatGPT responses, and transcribing Chat-GPT responses. Each task spans six cognitive processes defined in Bloom’s Taxonomy (remember, understand, apply, analyze, evaluate, and create). We extract interpretable temporal and rhythmic features and evaluate multiple classifiers under both Cognition-Aware and Cognition-Unaware settings. Temporal features perform well under Cognition-Aware evaluation scenarios, while rhythmic features generalize better under cross-cognition scenarios. Moreover, detecting bona fide and transcribed responses was easier than paraphrased ones for both the proposed models and human evaluators, with the models significantly outperforming the humans. Our findings affirm that keystroke dynamics facilitate reliable detection of LLM-assisted writing across varying cognitive demands and writing strategies, including paraphrasing and transcribing LLM-generated responses.

Keystroke dynamics is a promising modality for active user authentication, but its effectiveness under varying LLM-assisted typing and cognitive conditions remains understudied. Using data from 50 users and cognitive labels from Bloom’s Taxonomy, we evaluate keystroke-based authentication in Korean across three realistic typing scenarios: bona fide composition, LLM content paraphrasing, and transcription. Our pipeline incorporates continuity-aware segmentation, feature extraction, and classification via SVM, MLP, and XGB. Results show that the system maintains reliable performance across varying LLM usages and cognitive contexts, with Equal Error Rates ranging from 5.1% to 10.4%. These findings demonstrate the feasibility of behavioral authentication under modern writing conditions and offer insights into designing more context-resilient models.

Smartphones balance security and convenience by offering both knowledge-based (PINs, patterns) and biometric (facial, fingerprint) verification methods. However, studies have reported that PINs and patterns can be readily circumvented, while synthetically manipulated face data can easily deceive smartphone facial verification mechanisms. In this paper, we design a bimodal user verification mechanism that combines behavioral (pickup gesture) and biological (face) biometrics for user verification on smartphones. This work establishes a baseline for single-user verification scenarios on smartphones using a one-class verification model. The evaluation is performed in two stages: first, performance is assessed in both unimodal and bimodal settings using publicly available datasets; second, the robustness of the employed biological and behavioral traits is examined against four diverse attacks. Our findings emphasize the necessity of investigating diverse attack vectors, particularly fully synthetic data, to design robust user verification mechanisms.

Spotting and removing fake profiles could curb the menace of fake news in society. This paper, thus, investigates fake profile detection in social networks via users' typing patterns. We created a novel dataset of 468 posts from 26 users on three social networks: Facebook, Instagram, and$X$(previously Twitter) over six sessions. Then, we extract a series of features from keystroke timings and use them to predict whether two posts originated from the same users using three prominent statistical methods and their score-level fusion. The models' performance is evaluated under same, cross, and combined-cross-platform scenarios. We report the performance using k-rank accuracy for$k$varying from 1 to 5. The best-performing model obtained accuracies between 91.6%-100% on Facebook (Fusion), 70.8 - 87.5% on Instagram (Fusion), and 75% - 87.5% on X (Fusion) for$k$from 1 to 5. Under a cross-platform scenario, the fusion model achieved mean accuracies of 79.1% - 91.6%, 87.5% - 91.6%, and 83.3% - 87.5% when trained on Facebook, Instagram, and Twitter posts, respectively. In combined cross-platform, which involved mixing two platforms' data for model training while testing happened on the third platform's data, the best model achieved accuracy ranges of 75% - 95.8% across different scenarios. The results highlight the potential of the presented method in uncovering fake profiles across social network platforms.11©2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. To appear in 2024 IEEE 21st Consumer Communications & Networking Conference (CCNC)

The transition to online examinations and assignments raises significant concerns about academic integrity. Traditional plagiarism detection systems often struggle to identify instances of intelligent cheating, particularly when students utilize advanced generative AI tools to craft their responses. This study proposes a keystroke dynamics-based method to differentiate between bona fide and assisted writing within academic contexts. To facilitate this, a dataset was developed to capture the keystroke patterns of individuals engaged in writing tasks, both with and without the assistance of generative AI. The detector, trained using a modified TypeNet architecture, achieved accuracies ranging from 74.98% to 85.72% in condition-specific scenarios and 52.24% to 80.54% in condition-agnostic scenarios. The findings highlight significant differences in keystroke dynamics between genuine and assisted writing. The outcomes of this study enhance our understanding of how users interact with generative AI and have implications for improving the reliability of digital educational platforms.

This study investigates the vulnerabilities of data-driven offline signature verification (DASV) systems to generative attacks and proposes robust countermeasures. Specifically, we explore the efficacy of Variational Autoencoders (VAEs) and Conditional Generative Adversarial Networks (CGANs) in creating deceptive signatures that challenge DASV systems. Using the Structural Similarity Index (SSIM) to evaluate the quality of forged signatures, we assess their impact on DASV systems built with Xception, ResNet152V2, and DenseNet201 architectures. Initial results showed False Accept Rates (FARs) ranging from 0% to 5.47% across all models and datasets. However, exposure to synthetic signatures significantly increased FARs, with rates ranging from 19.12% to 61.64%. The proposed countermeasure, i.e., retraining the models with real + synthetic datasets, was very effective, reducing FARs between 0% and 0.99%. These findings emphasize the necessity of investigating vulnerabilities in security systems like DASV and reinforce the role of generative methods in enhancing the security of data-driven systems.

Integrating human feedback in models can improve the performance of natural language processing (NLP) models.Feedback can be either explicit (e.g.ranking used in training language models) or implicit (e.g. using human cognitive signals in the form of eyetracking).Prior eye tracking and NLP research reveal that cognitive processes, such as human scanpaths, gleaned from human gaze patterns aid in the understanding and performance of NLP models.However, the collection of real eyetracking data for NLP tasks is challenging due to the requirement of expensive and precise equipment coupled with privacy invasion issues.To address this challenge, we propose ScanTextGAN, a novel model for generating human scanpaths over text.We show that ScanTextGAN-generated scanpaths can approximate meaningful cognitive signals in human gaze patterns.We include synthetically generated scanpaths in four popular NLP tasks spanning six different datasets as proof of concept and show that the models augmented with generated scanpaths improve the performance of all downstream NLP tasks.

A recent study showed that commonly (vanilla) studied implementations of accelerometer-based gait authentication systems (vABGait) are susceptible to random-vector attack. The same study proposed a beta noise-assisted implementation (βABGait) to mitigate the attack. In this paper, we assess the effectiveness of the random-vector attack on both vABGait and (βABGait using three accelerometer-based gait datasets. In addition, we propose iABGait, an alternative implementation of ABGait, which uses a Conditional Tabular Generative Adversarial Network. Then we evaluate iABGait's resilience against the traditional zero-effort and random-vector attacks. The results show that iABGait mitigates the impact of the random-vector attack to a reasonable extent and outperforms βABGait in most experimental settings.

Previous studies have demonstrated that commonly studied (vanilla) touch-based continuous authentication systems (V-TCAS) are susceptible to population attack. This paper proposes a novel Generative Adversarial Network assisted TCAS (G-TCAS) framework, which showed more resilience to the population attack. G-TCAS framework was tested on a dataset of 117 users who interacted with a smartphone and tablet pair. On average, the increase in the false accept rates (FARs) for V-TCAS was much higher (22%) than G-TCAS (13%) for the smartphone. Likewise, the increase in the FARs for V-TCAS was 25% compared to G-TCAS (6%) for the tablet.

In this paper, we propose a novel continuous authentication system for smartphone users. The proposed system entirely relies on unlabeled phone movement patterns collected through smartphone accelerometer. The data was collected in a completely unconstrained environment over five to twelve days. The contexts of phone usage were identified using k-means clustering. Multiple profiles, one for each context, were created for every user. Five machine learning algorithms were employed for classification of genuine and impostors. The performance of the system was evaluated over a diverse population of 57 users. The mean equal error rates achieved by Logistic Regression, Neural Network, kNN, SVM, and Random Forest were 13.7%, 13.5%, 12.1%, 10.7%, and 5.6% respectively. A series of statistical tests were conducted to compare the performance of the classifiers. The suitability of the proposed system for different types of users was also investigated using the failure to enroll policy.

Despite the tremendous amount of research fronting the use of touch gestures as a mechanism of continuous authentication on smart phones, very little research has been conducted to evaluate how these systems could behave if attacked by sophisticated adversaries. In this article, we present two Lego-driven robotic attacks on touch-based authentication: a population statistics--driven attack and a user-tailored attack. The population statistics--driven attack is based on patterns gleaned from a large population of users, whereas the user-tailored attack is launched based on samples stolen from the victim. Both attacks are launched by a Lego robot that is trained on how to swipe on the touch screen. Using seven verification algorithms and a large dataset of users, we show that the attacks cause the system’s mean false acceptance rate (FAR) to increase by up to fivefold relative to the mean FAR seen under the standard zero-effort impostor attack. The article demonstrates the threat that robots pose to touch-based authentication and provides compelling evidence as to why the zero-effort attack should cease to be used as the benchmark for touch-based authentication systems.

Research on attacks which exploit video-based side-channels to decode text typed on a smartphone has traditionally assumed that the adversary is able to leverage some information from the screen display (say, a reflection of the screen or a low resolution video of the content typed on the screen). This paper introduces a new breed of side-channel attack on the PIN entry process on a smartphone which entirely relies on the spatio-temporal dynamics of the hands during typing to decode the typed text. Implemented on a dataset of 200 videos of the PIN entry process on an HTC One phone, we show, that the attack breaks an average of over 50% of the PINs on the first attempt and an average of over 85% of the PINs in ten attempts. Because the attack can be conducted in such a way not to raise suspicion (i.e., since the adversary does not have to direct the camera at the screen), we believe that it is very likely to be adopted by adversaries who seek to stealthily steal sensitive private information. As users conduct more and more of their computing transactions on mobile devices in the open, the paper calls for the community to take a closer look at the risks posed by the now ubiquitous camera-enabled devices.