Malware is becoming increasingly complex and widespread, making it essential to develop more effective and timely detection methods. Traditional static analysis often fails to defend against modern threats that employ code obfuscation, polymorphism, and other evasion techniques. In contrast, behavioral malware detection, which monitors runtime activities, provides a more reliable and context-aware solution. In this work, we propose BEACON, a novel deep learning framework that leverages large language models (LLMs) to generate dense, contextual embeddings from raw sandbox-generated behavior reports. These embeddings capture semantic and structural patterns of each sample and are processed by a one-dimensional convolutional neural network (1D CNN) for multi-class malware classification. Evaluated on the Avast-CTU Public CAPE Dataset, our framework consistently outperforms existing methods, highlighting the effectiveness of LLM-based behavioral embeddings and the overall design of BEACON for robust malware classification.

Approximate nearest neighbor search (ANNS) is broadly adopted in numerous scenarios. Real-world applications seek efficient ways to search billion-scale vectors in high throughput. On-SSD graph-based ANNS systems have the opportunity to achieve this goal, but the limited CPU computing power becomes a bottleneck. In this paper, we propose a GPU-centric, CPU-assisted ANNS architecture and design GustANN, a billion-scale graph-based vector search system for high throughput and cost-effectiveness. We achieve these goals with three techniques: (1) memory-efficient GPU kernels optimized to minimize the GPU memory usage in the graph search, which allows higher concurrency for GPU and SSD; (2) CPU-assisted transfer to address the PCIe bandwidth bottleneck on the GPU-side; (3) pivot search for inter-SSD load balancing. Compared to existing ANNS systems, GustANN achieves at least 2.50× higher throughput, and is 2.62× more cost-effective (measured in /QPS).

Deep learning recommendation models (DLRMs) have gained widespread application across search, advertising, and e-commerce. Still, DLRMs present notable challenges as they depend heavily on large embedding tables to represent sparse features in recommendation systems. This raises concerns about both memory capacity and cost. Solid-state drives (SSDs) offer a cost-effective solution with a significantly larger capacity, but they introduce read amplification issues because of the mismatch between embedding size and SSD read granularity. Prior SSD embedding storage systems aim to tackle these challenges by employing hypergraph partitioning to co-locate co-appearing embeddings onto the same SSD page, alleviating read amplification. However, this approach has a drawback as it divides embeddings into completely disjoint clusters, limiting potential combinations between embeddings.

With recent missions such as advanced space-based observatories like the Solar Dynamics Observatory (SDO) and Parker Solar Probe, and ground-based telescopes like the Daniel K. Inouye Solar Telescope (DKIST), the volume, velocity, and variety of data have made solar physics enter a transformative era as solar physics big data (SPBD). With the recent advancement of deep computer vision, there are new opportunities in SPBD for tackling previously unsolvable problems. However, new challenges arise due to the inherent characteristics of SPBD and deep computer vision models. This vision paper presents an overview of the different types of SPBD, explores new opportunities in applying deep computer vision to SPBD, highlights the unique challenges, and outlines several potential future research directions.

Solar activities are caused by the evolution of solar magnetic fields. Magnetic field parameters derived from photo-spheric vector magneto grams of solar active regions have been used to analyze and forecast extreme space weather events such as flares and coronal mass ejections. Unfortunately, the most recent Solar Cycle 24 was relatively weak with few large events, though it is the only solar cycle in which time-series vector magnetograms have been available. In this paper, we focus on two NASA instru-ments, namely the Michelson Doppler Imager (MDI) onboard the Solar and Heliospheric Observatory (SOHO) launched in Solar Cycle 23 (1996–2008), and the Helioseismic and Magnetic Imager (HMI) onboard the Solar Dynamics Observatory (SDO) launched in Solar Cycle 24 (2008–2019). While SOHOIMDI provides data from the more active Solar Cycle 23, it only offers line-of-sight (LOS) magneto grams without vector magnetograms. We propose Solar Image GAN (SIGAN), a generative adversarial network model designed to synthesize vector magnetic field images for Solar Cycles 23 and 24. SIGAN is trained using Hα images, SDOIHMI LOS, and vector magnetograms. It can generate vector magneto grams for both SDOIHMI and SOHOIMDI using Hα images and LOS magneto grams as input. Extensive experiments demonstrated the good performance of the proposed approach.

Deep learning models are often considered black box models as their internal workings tend to be opaque to the user. Because of this lack of transparency, it is challenging to understand the reasoning behind the model’s predictions. Here, we present an approach to making a solar flare prediction model interpretable. This model, built based on a long short-term memory (LSTM) network with an attention mechanism, aims to predict whether an active region (AR) on the Sun’s surface would produce a large flare, namely an M- or X-class flare, within 24 hours. The flare events used in this study are collected from the Geostationary Operational Environmental Satellite X-ray flare catalogs provided by the National Centers for Environmental Information. The crux of our approach is to model data samples in an AR as time series and use the LSTM network to capture the temporal dynamics of the data samples. Each data sample has 22 features including magnetic parameters and flare history parameters. To make the model’s predictions accountable and reliable, we leverage post hoc model-agnostic techniques, which help elucidate the factors contributing to the predicted output for an input sequence and provide insights into the model’s behavior across multiple sequences within an AR. To our knowledge, this is the first time that interpretability has been added to an LSTM-based flare prediction model.

In this paper we present a new approach, named DLGraph, for malware detection using deep learning and graph embedding. DLGraph employs two stacked denoising autoencoders (SDAs) for representation learning, taking into consideration computer programs' function-call graphs and Windows application programming interface (API) calls. Given a program, we first use a graph embedding technique that maps the program's function-call graph to a vector in a low-dimensional feature space. One SDA in our deep learning model is used to learn a latent representation of the embedded vector of the function-call graph. The other SDA in our model is used to learn a latent representation of the given program's Windows API calls. The two learned latent representations are then merged to form a combined feature vector. Finally, we use softmax regression to classify the combined feature vector for predicting whether the given program is malware or not. Experimental results based on different datasets demonstrate the effectiveness of the proposed approach and its superiority over a related method.

Directed networks find many applications in computer science, social science and biomedicine, among others. In this paper we propose a new graph mining algorithm that is capable of locating all frequent induced subgraphs in a given set of directed networks. We present an incremental coding scheme f or representing the canonical form of a graph, study its properties, and develop new techniques for pattern generation suitable for directed networks. We prove that our algorithm is complete, meaning that no qualified pattern is missed by the algorithm. Furthermore, our algorithm is correct in the sense that all patterns found by the algorithm are frequent induced subgraphs in the given networks. Experimental results based on synthetic data and gene regulatory networks show the good performance of our algorithm, and its application in network inference.

In systems biology, two important regulatory networks are gene regulatory networks (GRNs) and regulatory networks of microRNAs (RNMs). A GRN is modeled as a directed graph in which a node represents a gene or transcription factor (TF), and an edge from a TF to a gene indicates that the TF regulates the expression of the gene. An RNM is modeled as a bipartite directed graph with two disjoint sets of nodes: a set of nodes that represent microRNAs (miRNAs) and a set of nodes that represent genes or TFs. Directed edges between these two sets of nodes represent miRNA-target interactions or TF-miRNA regulatory relations. In this paper, we present an approach to reverse engineering GRNs and RNMs using a dynamic Bayesian network and mutual information scoring function. Our approach is able to automatically infer both GRNs and RNMs from time series of expression data. Experimental results on different datasets show that our approach is more accurate than other time-series based network inference methods.