Search

Foundation models refer to artificial intelligence (AI) models that are trained on massive amounts of data and demonstrate broad generalizability across various tasks with high accuracy. These models offer versatile, one-for-many or one-for-all solutions, eliminating the need for developing task-specific AI models. Examples of such foundation models include the Chat Generative Pre-trained Transformer (ChatGPT) and the Segment Anything Model (SAM). These models have been trained on millions to billions of samples and have shown wide-ranging and accurate applications in numerous tasks such as text processing (using ChatGPT) and natural image segmentation (using SAM). In medical image segmentation - finding target regions in medical images - there is a growing need for these one-for-many or one-for-all foundation models. Such models could obviate the need to develop thousands of task-specific AI models, which is currently standard practice in the field. They can also be adapted to tasks with datasets too small for effective training. We discuss two paths to achieve foundation models for medical image segmentation and comment on progress, challenges, and opportunities. One path is to adapt or fine-tune existing models, originally developed for natural images, for use with medical images. The second path entails building models from scratch, exclusively training on medical images.

As network services progress and mobile and IoT environments expand, numerous security concerns have surfaced for spectrum access systems. The omnipresent risk of Denial-of-Service (DoS) attacks and raising concerns about user privacy (e.g., location privacy, anonymity) are among such cyber threats. These security and privacy risks increase due to the threat of quantum computers that can compromise long-term security by circumventing conventional cryptosystems and increasing the cost of countermeasures. While some defense mechanisms exist against these threats in isolation, there is a significant gap in the state of the art on a holistic solution against DoS attacks with privacy and anonymity for spectrum management systems, especially when post-quantum (PQ) security is in mind. In this paper, we propose a new cybersecurity framework PACDoSQ, which is (to the best of our knowledge) the first to offer location privacy and anonymity for spectrum management with counter DoS and PQ security simultaneously. Our solution introduces the private spectrum bastion (database) concept to exploit existing architectural features of spectrum management systems and then synergizes them with multi-server private information retrieval and PQ-secure Tor to guarantee a location-private and anonymous acquisition of spectrum information together with hash-based client-server puzzles for counter DoS. We prove that PACDoSQ achieves its security objectives, and show its feasibility via a comprehensive performance evaluation., Comment: Accepted at IEEE Military Communications Conference, 28 October 1 November 2024, Washington, DC, USA

Recent breakthroughs in large language models (LLMs) offer unprecedented natural language understanding and generation capabilities. However, existing surveys on LLMs in biomedicine often focus on specific applications or model architectures, lacking a comprehensive analysis that integrates the latest advancements across various biomedical domains. This review, based on an analysis of 484 publications sourced from databases including PubMed, Web of Science, and arXiv, provides an in-depth examination of the current landscape, applications, challenges, and prospects of LLMs in biomedicine, distinguishing itself by focusing on the practical implications of these models in real-world biomedical contexts. Firstly, we explore the capabilities of LLMs in zero-shot learning across a broad spectrum of biomedical tasks, including diagnostic assistance, drug discovery, and personalized medicine, among others, with insights drawn from 137 key studies. Then, we discuss adaptation strategies of LLMs, including fine-tuning methods for both uni-modal and multi-modal LLMs to enhance their performance in specialized biomedical contexts where zero-shot fails to achieve, such as medical question answering and efficient processing of biomedical literature. Finally, we discuss the challenges that LLMs face in the biomedicine domain including data privacy concerns, limited model interpretability, issues with dataset quality, and ethics due to the sensitive nature of biomedical data, the need for highly reliable model outputs, and the ethical implications of deploying AI in healthcare. To address these challenges, we also identify future research directions of LLM in biomedicine including federated learning methods to preserve data privacy and integrating explainable AI methodologies to enhance the transparency of LLMs.

Given the rise in cyber threats to networked systems, coupled with the proliferation of AI techniques and enhanced processing capabilities, Denial of Service (DoS) attacks are becoming increasingly sophisticated and easily executable. They target system availability, compromising entire systems without breaking underlying security protocols. Consequently, numerous studies have focused on preventing, detecting, and mitigating DoS attacks. However, state-of-the-art systematization efforts have limitations such as isolated DoS countermeasures, shortcomings of AI-based studies, and a lack of DoS integration features like privacy, anonymity, authentication, and transparency. Additionally, the emergence of quantum computers is a game changer for DoS from attack and defense perspectives, yet it has remained largely unexplored. This study aims to address these gaps by examining (counter)-DoS in the AI era while also considering post-quantum (PQ) security when it applies. We highlight the deficiencies in the current literature and provide insights into synergistic techniques to bridge these gaps. We explore AI mechanisms for DoS intrusion detection, evaluate cybersecurity properties in cutting-edge machine learning models, and analyze weaponized AI in the context of DoS. We also investigate collaborative and distributed counter-DoS frameworks via federated learning and blockchains. Finally, we assess proactive approaches such as honeypots, puzzles, and authentication schemes that can be integrated into next-generation network systems for DoS prevention and mitigation., Comment: 10 Pages, 1 Figure, 2 Tables

Seismic wave amplification due to localized site conditions is an important aspect of regional seismic hazard assessment. Without systematic studies of frequency-dependent site-effects during strong Icelandic earthquakes, various local site proxies of large-scale studies in other seismic regions have been used in Iceland. Recently, earthquake site-effects were rigorously quantified for 34 stations in Southwest Iceland for the first time and correlated to distinct Icelandic geological units of hard rock, rock, lava rock, and sedimentary soil. These units are prevalent throughout Iceland and herein we present 1) nationwide maps of proxies (slope, Vs30, geological units) that may contribute to a better estimation of site effects and associated, 2) frequency-dependent site-amplification maps of Iceland. The frequency-dependent site factors for each geological unit are presented at 1-30 Hz and PGA. Finally, we generate site amplification maps based on recent large-scale models developed in other seismic regions (ESRM20) and various site proxies they are based on (geology- and slope-based inferred Vs30, geomorphological sedimentary thickness). We compare site-proxy maps and amplification maps from both Icelandic and large-scale, non-Icelandic, models. Neither spatial patterns nor amplification levels in either proxy or amplification maps from large-scale non-Icelandic studies resemble those observed from local quantitative strong-motion research as presented in this study. We attribute the discrepancy primarily to the young geology of Iceland and its formation history. Additionally, we compare model performance across frequencies by assessing the bias of model predictions against empirical site amplifications in the South Iceland Seismic Zone, accounting for site-to-site variability of residuals indicating the superior performance of the local amplification model.

Advancements in communication and information tech birthed the Smart Grid, optimizing energy and data transmission. Yet, user privacy is at risk due to frequent data collection. Existing privacy schemes face vulnerability with quantum machines. To tackle this, the LPM2DA scheme is introduced, utilizing lattice-based encryption and signatures for secure data aggregation. It ensures privacy, integrity, and authentication, enabling statistical analysis while preserving user privacy. Traditional aggregation schemes suffer from weak network models and centralization issues. Enter SPDBlock, a blockchain-based solution ensuring privacy, integrity, and resistance to attacks. It detects and prosecutes malicious entities while efficiently handling multi-dimensional data transmission. Through distributed decryption and secret sharing, only valid data can be decrypted with minimal involvement from smart meters. Performance tests reveal SPDBlock's superiority in communication and computational efficiency over traditional schemes., Comment: 103 pages, 8 figures

COVID-19 has deeply affected daily life and travel behaviors. Understanding these changes is crucial, prompting an investigation into socio-demographic and socio-economic factors. This study used large-scale mobile device location data in Washington, D.C., Maryland, and Virginia (DMV area) to unveil the impacts of these variables on commute trip changes. It reflected short and long-term impacts through linear regression and geographically weighted regression models. Findings indicated that counties with a higher percentage of people using walking and biking during the initial phase of COVID-19 experienced greater reductions in commute trips. For the long-term effect in November, the impact of active modes became insignificant, and individuals using public modes showed more significant trip reductions. Positive correlations were observed between median income levels and reduced commute trips. Sectors requiring ongoing outdoor operations during the pandemic showed substantial negative correlations. In the DMV area, counties with a higher proportion of Democratic voters experienced less trip reduction. Applying Geographically Weighted Regression models captured local spatial relationships, showing the emergence of local correlations as the pandemic evolved, suggesting a geographical impact pattern. Initially global, the pandemic's impact on commuting behaviors became more influenced by spatial factors over time, showing localized effects., Comment: 28 pages, 8 figures, accepted at TRR 2024

The Medical Internet of Things (MIoT) harbors resource-limited medical embedded devices that collect security-sensitive data from users for analysis, monitoring, and diagnosis. Digital signatures play a foundational role in ensuring the authentication and integrity of this sensitive medical information, critical for the trustworthiness of MIoT applications. However, traditional signatures used in current IoT systems may lack the necessary long-term security and are vulnerable to emerging quantum computer threats. NISt's PQC standards impose heavy overhead unsuitable for battery-limited MIoT devices. Efforts to design more computationally efficient PQ signatures have faced challenges, either introducing significant memory overhead and potential vulnerabilities or relying on strong assumptions. Hence, this paper introduces INFinity-HORS (INF-HORS), a lightweight PQ digital signature. To the best of our knowledge, INF-HORS is the first signer-optimal hash-based signature offering polynomial unbounded signing capabilities under minimal architectural assumptions. Unlike other PQ signatures, INF-HORS does not require hyper-tree structures or incur the high memory usage seen in multivariate counterparts. Our performance analysis confirms that INF-HORS is significantly more computationally efficient than NIST PQC standards like Dilithium and SPHINCS+. We prove INF-HORS's security in the random oracle model and show through experiments that it achieves 20x faster signature generation and smaller signature and private key sizes compared to BLISS-I on an 8-bit ATxmega128A1 microcontroller. INF-HORS does not rely on non-colluding verification servers, secure enclaves, or trusted verification assisting entities, minimizing security risks and making it ideal for MIoT with minimal cryptographic overhead and strong security assumptions., Comment: 14 pages

In this paper, we delve into the susceptibility of federated medical image analysis systems to adversarial attacks. Our analysis uncovers a novel exploitation avenue: using gradient information from prior global model updates, adversaries can enhance the efficiency and transferability of their attacks. Specifically, we demonstrate that single-step attacks (e.g. FGSM), when aptly initialized, can outperform the efficiency of their iterative counterparts but with reduced computational demand. Our findings underscore the need to revisit our understanding of AI security in federated healthcare settings.

The rise of quantum computers exposes vulnerabilities in current public key cryptographic protocols, necessitating the development of secure post-quantum (PQ) schemes. Hence, we conduct a comprehensive study on various PQ approaches, covering the constructional design, structural vulnerabilities, and offer security assessments, implementation evaluations, and a particular focus on side-channel attacks. We analyze global standardization processes, evaluate their metrics in relation to real-world applications, and primarily focus on standardized PQ schemes, selected additional signature competition candidates, and PQ-secure cutting-edge schemes beyond standardization. Finally, we present visions and potential future directions for a seamless transition to the PQ era.

Optimization-based regularization methods have been effective in addressing the challenges posed by data heterogeneity in medical federated learning, particularly in improving the performance of underrepresented clients. However, these methods often lead to lower overall model accuracy and slower convergence rates. In this paper, we demonstrate that using Vision Transformers can substantially improve the performance of underrepresented clients without a significant trade-off in overall accuracy. This improvement is attributed to the Vision transformer's ability to capture long-range dependencies within the input data.

This paper explores the security aspects of federated learning applications in medical image analysis. Current robustness-oriented methods like adversarial training, secure aggregation, and homomorphic encryption often risk privacy compromises. The central aim is to defend the network against potential privacy breaches while maintaining model robustness against adversarial manipulations. We show that incorporating distributed noise, grounded in the privacy guarantees in federated settings, enables the development of a adversarially robust model that also meets federated privacy standards. We conducted comprehensive evaluations across diverse attack scenarios, parameters, and use cases in cancer imaging, concentrating on pathology, meningioma, and glioma. The results reveal that the incorporation of distributed noise allows for the attainment of security levels comparable to those of conventional adversarial training while requiring fewer retraining samples to establish a robust model.

Federated learning offers a privacy-preserving framework for medical image analysis but exposes the system to adversarial attacks. This paper aims to evaluate the vulnerabilities of federated learning networks in medical image analysis against such attacks. Employing domain-specific MRI tumor and pathology imaging datasets, we assess the effectiveness of known threat scenarios in a federated learning environment. Our tests reveal that domain-specific configurations can increase the attacker's success rate significantly. The findings emphasize the urgent need for effective defense mechanisms and suggest a critical re-evaluation of current security protocols in federated medical image analysis systems.

In recent years, passively collected GPS data have been popularly applied in various transportation studies, such as highway performance monitoring, travel behavior analysis, and travel demand estimation. Despite multiple advantages, one of the issues is data oscillations (aka outliers or data jumps), which are unneglectable since they may distort mobility patterns and lead to wrongly or biased conclusions. For transportation studies driven by GPS data, assuring the data quality by removing noises caused by data oscillations is undoubtedly important. Most GPS-based studies simply remove oscillations by checking the high speed. However, this method can mistakenly identify normal points as oscillations. Some other studies specifically discuss the removal of outliers in GPS data, but they all have limitations and do not fit passively collected GPS data. Many studies are well developed for addressing the ping-pong phenomenon in cellular data, or cellular tower data, but the oscillations in passively collected GPS data are very different for having much more various and complicated patterns and being more uncertain. Current methods are insufficient and inapplicable to passively collected GPS data. This paper aims to address the oscillated points in passively collected GPS data. A set of heuristics are proposed by identifying the abnormal movement patterns of oscillations. The proposed heuristics well fit the features of passively collected GPS data and are adaptable to studies of different scales, which are also computationally cost-effective in comparison to current methods.