arXiv Computer Science @arxiv_cs@qoto.org

Healthcare decision-making represents one of the most challenging domains for Artificial Intelligence (AI), requiring the integration of diverse knowledge sources, complex reasoning, and various external analytical tools. Current AI systems often rely on either task-specific models, which offer limited adaptability, or general language models without grounding with specialized external knowledge and tools. We introduce MedOrch, a novel framework that orchestrates multiple specialized tools and reasoning agents to provide comprehensive medical decision support. MedOrch employs a modular, agent-based architecture that facilitates the flexible integration of domain-specific tools without altering the core system. Furthermore, it ensures transparent and traceable reasoning processes, enabling clinicians to meticulously verify each intermediate step underlying the system's recommendations. We evaluate MedOrch across three distinct medical applications: Alzheimer's disease diagnosis, chest X-ray interpretation, and medical visual question answering, using authentic clinical datasets. The results demonstrate MedOrch's competitive performance across these diverse medical tasks. Notably, in Alzheimer's disease diagnosis, MedOrch achieves an accuracy of 93.26%, surpassing the state-of-the-art baseline by over four percentage points. For predicting Alzheimer's disease progression, it attains a 50.35% accuracy, marking a significant improvement. In chest X-ray analysis, MedOrch exhibits superior performance with a Macro AUC of 61.2% and a Macro F1-score of 25.5%. Moreover, in complex multimodal visual question answering (Image+Table), MedOrch achieves an accuracy of 54.47%. These findings underscore MedOrch's potential to advance healthcare AI by enabling reasoning-driven tool utilization for multimodal medical data processing and supporting intricate cognitive tasks in clinical decision-making.

Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, offer compact and effective alternatives to full model fine-tuning by introducing low-rank updates to pretrained weights. However, most existing approaches rely on global low-rank structures, which can overlook spatial patterns spread across the parameter space. In this work, we propose Localized LoRA, a generalized framework that models weight updates as a composition of low-rank matrices applied to structured blocks of the weight matrix. This formulation enables dense, localized updates throughout the parameter space-without increasing the total number of trainable parameters. We provide a formal comparison between global, diagonal-local, and fully localized low-rank approximations, and show that our method consistently achieves lower approximation error under matched parameter budgets. Experiments on both synthetic and practical settings demonstrate that Localized LoRA offers a more expressive and adaptable alternative to existing methods, enabling efficient fine-tuning with improved performance.

Natural disasters usually affect vast areas and devastate infrastructures. Performing a timely and efficient response is crucial to minimize the impact on affected communities, and data-driven approaches are the best choice. Visual question answering (VQA) models help management teams to achieve in-depth understanding of damages. However, recently published models do not possess the ability to answer open-ended questions and only select the best answer among a predefined list of answers. If we want to ask questions with new additional possible answers that do not exist in the predefined list, the model needs to be fin-tuned/retrained on a new collected and annotated dataset, which is a time-consuming procedure. In recent years, large-scale Vision-Language Models (VLMs) have earned significant attention. These models are trained on extensive datasets and demonstrate strong performance on both unimodal and multimodal vision/language downstream tasks, often without the need for fine-tuning. In this paper, we propose a VLM-based zero-shot VQA (ZeShot-VQA) method, and investigate the performance of on post-disaster FloodNet dataset. Since the proposed method takes advantage of zero-shot learning, it can be applied on new datasets without fine-tuning. In addition, ZeShot-VQA is able to process and generate answers that has been not seen during the training procedure, which demonstrates its flexibility.

The ability of AI to sense and identify various substances based on their smell alone can have profound impacts on allergen detection (e.g., smelling gluten or peanuts in a cake), monitoring the manufacturing process, and sensing hormones that indicate emotional states, stress levels, and diseases. Despite these broad impacts, there are virtually no large scale benchmarks, and therefore little progress, for training and evaluating AI systems' ability to smell in the real world. In this paper, we use portable gas and chemical sensors to create SmellNet, the first large-scale database that digitizes a diverse range of smells in the natural world. SmellNet contains about 180,000 time steps of 50 substances (spanning nuts, spices, herbs, fruits, and vegetables) with 50 hours of data. Using SmellNet, we train AI models for real-time classification of substances based on their smell alone. Our best methods leverage sequence models, contrastive learning to integrate high-resolution Gas Chromatography-Mass Spectrometry molecular data, and a new temporal difference method that identifies sharp changes in sensor readings. Our best models achieve up to 65.35% accuracy on pre-recorded data, and generalize to real-world conditions with 10.71% accuracy on nuts and 25.38% on spices in the challenging 50-way online classification task. Despite these promising results, SmellNet highlights many technical challenges in building AI for smell, including richer feature learning, on-edge smell models, and robustness to environmental changes.

Since the 1990s, the integration of technology into daily life has led to the creation of an extensive network of interconnected devices, transforming how individuals and organizations operate. However, this digital transformation has also spurred the rise of cybercrime, criminal activities perpetrated through networks or computer systems. Cybercrime has become a global concern, presenting significant challenges to security systems. Although advancements in digital technology have enhanced efficiency, they have also opened new avenues for exploitation by cybercriminals, highlighting the urgent need for advanced cybersecurity measures. The escalating number of cyberattacks and associated risks in the past decade highlights the critical importance of protecting sensitive data and safeguarding information systems. Cybercrimes range from financial fraud and phishing scams to identity theft and online harassment, posing substantial risks to both individuals and organizations. In response, governments, law enforcement agencies, and cybersecurity units have intensified their efforts to address these threats. In recent years, India has experienced a significant surge in cybercrime incidents, with a notable increase in cases involving ransomware, data breaches, and social engineering attacks. The growing penetration of internet services, the expansion of e-commerce, and the rapid adoption of digital payment systems have made individuals and organizations more vulnerable to cyber threats. Key areas affected include banking, healthcare, and government sectors, which are frequently targeted due to the sensitive nature of the data they handle. To combat these risks, there is an increasing focus on public awareness, cybersecurity education, and robust regulatory frameworks. This paper examines cybercrime, prevention strategies, security protocols, and terminology to safeguard digital infrastructure.

Advanced Encryption Standard (AES) is one of the most widely used symmetric cipher for the confidentiality of data. Also it is used for other security services, viz. integrity, authentication and key establishment. However, recently, authors have shown some weakness in the generation of sub-keys in AES, e.g. bit leakage attack, etc. Also, AES sub-keys are generated sequentially, which is an overhead, especially for resource-constrained devices. Therefore, we propose and investigate a novel encryption AESHA3, which uses sub-keys generated by Secure Hash Algorithm-3 (SHA3). The output of SHA3 is one-way and highly non-linear, and random. The experimental analysis shows that the average time taken for generating the sub-keys to be used for encrypting the data using our approach i.e. AESHA3 is ~1300 times faster than the sub-key generated by the standard AES. Accordingly, we find that AESHA3 will be very relevant not only in terms of security but also it will save the resources in IoT devices. We investigated AESHA3 in Intel Core i7, 6th Generation processor and Raspberry Pi 4B and found that up to two MB data encryption is very significant, and lesser the data size, more the resource saving compared to AES.

Digital signature algorithms (DSAs) are fundamental to cryptographic security, ensuring data integrity and authentication. While RSA, DSA, ECDSA, and EdDSA are widely used, their performance varies significantly depending on key sizes, hash functions, and elliptic curve configurations. In this paper, we introduce LightDSA, a hybrid and configurable digital signature library that supports RSA, DSA, ECDSA, and EdDSA with flexible form and curve selection, open sourced at https://github.com/serengil/LightDSA. Unlike conventional implementations that impose strict curve-form mappings - such as Weierstrass for ECDSA and Edwards for EdDSA LightDSA - allows arbitrary combinations, enabling a broader performance evaluation. We analyze the computational efficiency of these algorithms across various configurations, comparing key generation, signing, and verification times. Our results provide insights into the trade-offs between security and efficiency, guiding the selection of optimal configurations for different cryptographic needs.