Search for: All records

Millions of new pieces of malicious software (i.e., malware) are introduced each year. This poses significant challenges for antivirus vendors, who use machine learning to detect and analyze malware, and must keep up with changes in the distribution while retaining knowledge of older variants. Continual learning (CL) holds the potential to address this challenge by relaxing the requirements of the incremental storage and computational costs of regularly retraining over all the collected data. Prior work, however, shows that CL techniques, which are designed primarily for computer vision tasks, fare poorly when applied to malware classification. To address these issues, we begin with an exploratory analysis of a typical malware dataset, which reveals that malware families are heterogeneous and difficult to characterize, requiring a wide variety of samples to learn a robust representation. Based on these findings, we propose Malware Analysis with Distribution-Aware Replay (MADAR), a CL framework that accounts for the unique properties and challenges of the malware data distribution. Through extensive evaluation on large-scale Windows and Android malware datasets, we show that MADAR significantly outperforms prior work. This highlights the importance of understanding domain characteristics when designing CL techniques and demonstrates a path forward for the malware analysis domain. 

This paper aims to study such relationships between models and datasets, which are the central parts of the LLM supply chain. 

This work introduces a novel physics-informed neural network (PINN)-based framework for modeling and optimizing false data injection (FDI) attacks on electric vehicle charging station (EVCS) networks, with a focus on centralized charging management system (CMS). By embedding the governing physical laws as constraints within the neural network’s loss function, the proposed framework enables scalable, real-time analysis of cyber-physical vulnerabilities. The PINN models EVCS dynamics under both normal and adversarial conditions while optimizing stealthy attack vectors that exploit voltage and current regulation. Evaluations on the IEEE 33-bus system demonstrate the framework’s capability to uncover critical vulnerabilities. These findings underscore the urgent need for enhanced resilience strategies in EVCS networks to mitigate emerging cyber threats targeting the power grid. Furthermore, the framework lays the groundwork for exploring a broader range of cyber-physical attack scenarios on EVCS networks, offering potential insights into their impact on power grid operations. It provides a flexible platform for studying the interplay between physical constraints and adversarial manipulations, enhancing our understanding of EVCS vulnerabilities. This approach opens avenues for future research into robust mitigation strategies and resilient design principles tailored to the evolving cybersecurity challenges in smart grid systems. 

Seismocardiography (SCG) has attracted significant interest for monitoring cardiac health and diagnosing cardiovascular conditions. While traditional SCG methods rely on uncomfortable chest-mounted accelerometers, recent research explores non-contact approaches, including analyzing video recordings of the chest. In this study, three computer vision-based methods including Lucas-Kanade optical flow, template tracking, and Gunnar-Farneback optical flow were evaluated for extracting SCG signals from ordinary camera-recorded chest videos. The study focused on right-to-left and head-to-foot SCG signals obtained from 13 healthy subjects during breath-hold at the end of exhalation and inhalation. Comparative analysis was performed by calculating the mean squared error (MSE) and root MSE (RMSE) between the vision-based SCG signals and the gold-standard accelerometer signals. Visual and quantitative analyses showed that the Lucas-Kanade and template tracking methods estimated vision-based SCG signals closely resembling the accelerometer data, particularly in the head-to-foot direction. The Lucas-Kanade method had MSE values ranging from 0.14 to 0.93, RMSE values from 0.38 to 0.96, average correlation values of 0.82±0.09. The template tracking method showed MSE values between 0.12 to 0.94, RMSE values from 0.35 to 0.97, and average correlation values of 0.83±0.10. In comparison, the Farneback method had higher MSE values ranging from 0.20 to 1.07, RMSE values from 0.44 to 1.03, and average correlation values of 0.76±0.11. These results suggest the effectiveness of Lucas-Kanade and template tracking methods for non-contact SCG signal extraction from chest video data. 

Introduction:Seismocardiography (SCG) - measurements of cardiovascular-induced vibrations on the chest - has shown potential for providing clinical information for cardiac conditions. SCG is conventionally recorded by an accelerometer attached to a single point on chest. Recent research suggests multichannel SCG (mSCG) - measurements from multiple chest locations - can provide extra and more accurate clinical information. Current mSCG methods are limited to accelerometer arrays, laser Doppler vibrometry, and airborne ultrasound that are either costly, difficult for inexperienced users, or need bulky equipment, thereby impeding their use beyond research or clinical settings. Hypothesis:mSCG signals can be accurately estimated from tiny chest movements in chest videos recorded by ordinary cameras, e.g., those in smartphones. Methods:We enrolled 10 subjects (sbjs) with no history of CVDs (21.7 ± 1.7 years, 40% women). ECG and chest video of sbjs were recorded at rest for 15 sec during breath hold at the end of inhalation followed by another 15 sec recording during breath hold at the end of exhalation. We developed an AI-powered mobile app to record the chest videos and convert them to 0-30 Hz mSCG in right-to-left (RL) and head-to-foot (HF) directions (Fig 1a). Heart rate (HR) based on ECG RR interval and mSCG was measured and compared. Results:HR estimated from mSCG in both RL and HF directions had a good agreement with ECG-based HR using Bland-Altman analysis [RL: bias = 1.4 bpm, 95% CI = 5.6 bpm; HF: bias = 0.8 bpm, 95% CI = 6.2 bpm (Fig 1b)]. High-quality mSCG and ECG measurements were obtained for all sbjs. Conclusion:Clinically relevant information can be accurately extracted from chest videos using our novel, contactless, AI-based method. Given that the vast majority of Americans have access to a camera phone, future developments of this method may provide new means of remote and accessible cardiac monitoring.