• A practical expert system for detecting audio deepfakes under low-label conditions. • Combines spectral-temporal graph construction with vision graph encoders. • Leverages label-free non-contrastive learning for robust audio representation. • Outperforms supervised and self-supervised baselines using only 5 labeled data. • Demonstrates strong generalization across attack types, languages, and domains. Audio deepfake detection is increasingly important as synthetic speech becomes more realistic and accessible. Recent methods, including those using graph neural networks (GNNs) to model frequency and temporal dependencies, show strong potential but need large amounts of labeled data, which limits their practical use. Label-efficient alternatives like graph-based non-contrastive learning offer a potential solution, as they can learn useful representations from unlabeled data without using negative samples. However, current graph non-contrastive approaches are built for single-view graph representations and cannot be directly used for audio, which has unique spectral and temporal structures. Bridging this gap requires dual-view graph modeling suited to audio signals. In this work, we introduce SIGNL (The code is available at https://github.com/falihgoz/SIGNL .) (Spectral-temporal vIsion Graph Non-contrastive Learning), a label-efficient expert system for detecting audio deepfakes. SIGNL operates on the visual representation of audio—such as spectrograms or other time-frequency encodings—transforming them into spectral and temporal graphs for structured feature extraction. It then employs graph convolutional encoders to learn complementary frequency-time features, effectively capturing the unique characteristics of audio. These encoders are pre-trained using a non-contrastive self-supervised learning strategy on augmented graph pairs, enabling effective representation learning without labeled data. The resulting encoders are then fine-tuned on minimal labelled data for downstream deepfake detection. SIGNL achieves strong performance on multiple audio deepfake detection benchmarks, including 7.88% EER on ASVspoof 2021 DF and 3.95% EER on ASVspoof 5 using only 5% labeled data. It also generalizes well to unseen conditions, reaching 10.16% EER on the In-The-Wild dataset when trained on CFAD.

The performance of existing audio deepfake detection frameworks degrades when confronted with new deepfake attacks. Rehearsal-based continual learning (CL), which updates models using a limited set of old data samples, helps preserve prior knowledge while incorporating new information. However, existing rehearsal techniques don't effectively capture the diversity of audio characteristics, introducing bias and increasing the risk of forgetting. To address this challenge, we propose Rehearsal with Auxiliary-Informed Sampling (RAIS), a rehearsal-based CL approach for audio deepfake detection. RAIS employs a label generation network to produce auxiliary labels, guiding diverse sample selection for the memory buffer. Extensive experiments show RAIS outperforms state-of-the-art methods, achieving an average Equal Error Rate (EER) of 1.953 % across five experiences. The code is available at: https://github.com/falihgoz/RAIS.

Psychiatric disorders involve complex neural activity changes, with functional magnetic resonance imaging (fMRI) data serving as key diagnostic evidence. However, data scarcity and the diverse nature of fMRI information pose significant challenges. While graph-based self-supervised learning (SSL) methods have shown promise in brain network analysis, they primarily focus on time-domain representations, often overlooking the rich information embedded in the frequency domain. To overcome these limitations, we propose F requency- E nhanced Net work (FENet), a novel SSL framework specially designed for fMRI data that integrates time-domain and frequency-domain information to improve psychiatric disorder detection in small-sample datasets. FENet constructs multi-view brain networks based on the inherent properties of fMRI data, explicitly incorporating frequency information into the learning process of representation. Additionally, it employs domain-specific encoders to capture temporal-spectral characteristics, including an efficient frequency-domain encoder that highlights disease-relevant frequency features. Finally, FENet introduces a domain consistency-guided learning objective, which balances the utilization of diverse information and generates frequency-enhanced brain graph representations. Experiments on two real-world medical datasets demonstrate that FENet outperforms state-of-the-art methods while maintaining strong performance in minimal data conditions. Furthermore, we analyze the correlation between various frequency-domain features and psychiatric disorders, emphasizing the critical role of high-frequency information in disorder detection.

Many multivariate time series anomaly detection frameworks have been proposed and widely applied. However, most of these frameworks do not consider intrinsic relationships between variables in multivariate time series data, thus ignoring the causal relationship among variables and degrading anomaly detection performance. This work proposes a novel framework called CGAD, an entropy causal graph for multivariate time series Anomaly Detection. CGAD utilizes transfer entropy to construct graph structures that unveil the underlying causal relationships among time series data. Weighted graph convolutional networks combined with causal convolutions are employed to model both the causal graph structures and the temporal patterns within multivariate time series data. Furthermore, CGAD applies anomaly scoring, leveraging median absolute deviation-based normalization to improve the robustness of the anomaly identification process. Extensive experiments demonstrate that CGAD outperforms state-of-the-art methods on real-world datasets with a 9% average improvement in terms of three different multivariate time series anomaly detection metrics.

System logs are used to record the operating system and application program runtime status, constituting a form of semi-structured text. Analyzing these logs becomes pivotal in addressing failures or abnormalities within the operating system or application. However, the substantial volume of log data necessitates extensive manual analysis. To overcome this problem, researchers propose an automated log parsing method that reduces the volume of logs by parsing them into templates. However, a profound understanding of these templates still requires domain-specific knowledge. Therefore, this paper proposes EntityParser, which aims at enhancing log semantics. EntityParser identifies entities within log templates, mapping abstract log messages to specific, meaningful entities, thereby enhancing the readability of the logs. Additionally, EntityParser maintains two high-quality knowledge bases and can enrich them through self-iteration.Finally, in order to validate the efficacy of EntityParser, we conduct experiments on the LogHub dataset and the Linux system log dataset collected from real production environments. The results demonstrate the effectiveness of the proposed method in improving log semantics, discovering new entities, and proving its generalization. Furthermore, it enables the discovery of all potential entities within the current log through self-iteration.

To exploit high temporal correlations in video frames of the same scene, the current frame is predicted from the already-encoded reference frames using block-based motion estimation and compensation techniques. While this approach can efficiently exploit the translation motion of the moving objects, it is susceptible to other types of affine motion and object occlusion/deocclusion. Recently, deep learning has been used to model the high-level structure of human pose in specific actions from short videos and then generate virtual frames in future time by predicting the pose using a generative adversarial network (GAN). Therefore, modelling the high-level structure of human pose is able to exploit semantic correlation by predicting human actions and determining its trajectory. Video surveillance applications will benefit as stored “big” surveillance data can be compressed by estimating human pose trajectories and generating future frames through semantic correlation. This paper explores a new way of video coding by modelling human pose from the already-encoded frames and using the generated frame at the current time as an additional forward-referencing frame. It is expected that the proposed approach can overcome the limitations of the traditional backward-referencing frames by predicting the blocks containing the moving objects with lower residuals. Our experimental results show that the proposed approach can achieve on average up to 2.83 dB PSNR gain and 25.93% bitrate savings for high motion video sequences compared to standard video coding.

Abstract The main aim of the outlying aspect mining algorithm is to automatically detect the subspace(s) (a.k.a. aspect(s)), where a given data point is dramatically different than the rest of the data in each of those subspace(s) (aspect(s)). To rank the subspaces for a given data point, a scoring measure is required to compute the outlying degree of the given data in each subspace. In this paper, we introduce a new measure to compute outlying degree, called Simple Isolation score using Nearest Neighbor Ensemble (SiNNE), which not only detects the outliers but also provides an explanation on why the selected point is an outlier. SiNNE is a dimensionally unbias measure in its raw form, which means the scores produced by SiNNE are compared directly with subspaces having different dimensions. Thus, it does not require any normalization to make the score unbiased. Our experimental results on synthetic and publicly available real-world datasets revealed that (i) SiNNE produces better or at least the same results as existing scores. (ii) It improves the run time of the existing outlying aspect mining algorithm based on beam search by at least two orders of magnitude. SiNNE allows the existing outlying aspect mining algorithm to run in datasets with hundreds of thousands of instances and thousands of dimensions which was not possible before.

Diabetic eye disease (DED) is a cluster of eye problem that affects diabetic patients. Identifying DED is a crucial activity in retinal fundus images because early diagnosis and treatment can eventually minimize the risk of visual impairment. The retinal fundus image plays a significant role in early DED classification and identification. An accurate diagnostic model's development using a retinal fundus image depends highly on image quality and quantity. This paper presents a methodical study on the significance of image processing for DED classification. The proposed automated classification framework for DED was achieved in several steps: image quality enhancement, image segmentation (region of interest), image augmentation (geometric transformation), and classification. The optimal results were obtained using traditional image processing methods with a new build convolution neural network (CNN) architecture. The new built CNN combined with the traditional image processing approach presented the best performance with accuracy for DED classification problems. The results of the experiments conducted showed adequate accuracy, specificity, and sensitivity.

Positive unlabeled time series classification (PUTSC) refers to classifying time series with a set PL of positive labeled examples and a set U of unlabeled ones. Model selection for PUTSC is a largely untouched topic. In this paper, we look into PUTSC model selection, which as far as we know is the first systematic study in this topic. Focusing on the widely adopted self-training one-nearest-neighbor (ST-1NN) paradigm, we propose a model selection framework based on active learning (AL). We present the novel concepts of self-training label propagation, pseudo label calibration principles and ultimately influence to fully exploit the mechanism of ST-1NN. Based on them, we develop an effective model performance evaluation strategy and three AL sampling strategies. Experiments on over 120 datasets and a case study in arrhythmia detection show that our methods can yield top performance in interactive environments, and can achieve near optimal results by querying very limited numbers of labels from the AL oracle.

Absence epilepsy is one of the most common types of epilepsy. The diagnosis of absence epilepsy is among the greatest challenges faced by clinical neurologists due to a lack of easily observable symptoms that are present in conventional epilepsy (e.g. spasm and convulsion), and highly relies on the detection of Spike and Slow Waves (SSWs) in Electroencephalogram (EEG) signals. Recently, graph representations called complex networks have been increasingly applied to characterizing 1D EEG signals. However, existing methods often fail to effectively represent SSWs, struggling to capture the differences between SSW waveforms and their non-SSW counterparts, such as minute differences and distinct shapes. Addressing this issue, in this work, we propose two simple yet effective complex networks, Overlook Graph (OG) and Weighted Overlook Graph (WOG), which have been customized to expressively represent SSWs. Built upon OG and WOG, we then develop a 2D Convolutional Neural Network (2D-CNN) to further learn latent features from the graph representations and accomplish the detection task. Extensive experiments on a real-world absence epilepsy EEG dataset show that the proposed OG/WOG-2D-CNN method can accurately detect SSWs. Additional experiments on the well-known Bonn dataset further show that our method can generalize to the conventional epilepsy seizure detection task with highly competitive performances.

Regularities analysis for prescriptions is a significant task for traditional Chinese medicine (TCM), both in inheritance of clinical experience and in improvement of clinical quality. Recently, many methods have been proposed for regularities discovery, but this task is challenging due to the quantity, sparsity and free-style of prescriptions. In this paper, we address the specific problem of regularities discovery and propose a graph embedding based framework for regularities discovery for massive prescriptions. We model this task as a relation prediction in which the correlation of two herbs or of herb and symptom are incorporated to characterize the different relationships. Specifically, we first establish a heterogeneous network with herbs and symptoms as its nodes. We develop a bipartite embedding model termed HS2Vec to detect regularities, which explores multiple relations of herbherb, and herb-symptom based on the heterogeneous network. Experiments on four real-world datasets demonstrate that the proposed framework is very effective for regularities discovery.

There has been a continuing demand for traditional and complementary medicine worldwide. A fundamental and important topic in Traditional Chinese Medicine (TCM) is to optimize the prescription and to detect herb regularities from TCM data. In this paper, we propose a novel clustering model to solve this general problem of herb categorization, a pivotal task of prescription optimization and herb regularities. The model utilizes Random Walks method, Bayesian rules and Expectation Maximization(EM) models to complete a clustering analysis effectively on a heterogeneous information network. We performed extensive experiments on the real-world datasets and compared our method with other algorithms and experts. Experimental results have demonstrated the effectiveness of the proposed model for discovering useful categorization of herbs and its potential clinical manifestations.

Recent technological advances have accelerated the design and deployment of kinds of secure applications on smartphones. Although users can access and handle their data flexibly and stably with mobile devices, not only computing devices, it poses security challenges of a new dimension that users disclose lots of sensitive data and privacy information over open devices and networks as well. Thus, more and more malwares are emerging to compromise mobile OS and steal sensitive data from these applications. In this paper, we propose a mobile payment framework TrustPAY on TrustZone security enhanced platform, which can ensure payment transactions security and realize privacy friendly payment. We have implemented a prototype system on a simulation environment by using ARM FastModel and Open Virtualization software stack for ARM TrustZone, and presented our implementation on a real development board by using ARM CoreTile Express A9×4. Our experiment evaluation and security analysis prove that our scheme can effectively meet the security requirements of a practical m-payment with acceptable performance. Furthermore, TrustPAY is also flexible to support kinds of secure applications requiring to privacy protection.

Uncertain data streams have been widely generated in many Web applications. The uncertainty in data streams makes anomaly detection from sensor data streams far more challenging. In this article, we present a novel framework that supports anomaly detection in uncertain data streams. The proposed framework adopts the wavelet soft-thresholding method to remove the noises or errors in data streams. Based on the refined data streams, we develop effective period pattern recognition and feature extraction techniques to improve the computational efficiency. We use classification methods for anomaly detection in the corrected data stream. We also empirically show that the proposed approach shows a high accuracy of anomaly detection on several real datasets.

With the advent of Cloud computing and subsequent big data, online decision makers usually find it difficult to make informed decisions because of the great amount of irrelevant, uncertain, or inaccurate information. In this paper, we explore the application of multicriteria decision-making (MCDM) techniques in the area of Cloud computing and big data, to find an efficient way of dealing with criteria relations and fuzzy knowledge based on a great deal of information. We propose a MCDM framework, which combines the ISM-based and ANP-based techniques, to model the interactive relations between evaluation criteria, and to handle data uncertainties. We present an application of Cloud service selection to prove the efficiency of the proposed framework, in which a user-oriented sigmoid utility function is designed to evaluate the performance of each criterion.

QoS-based service rating has made positive contributions to the area of service selection. Especially for Cloud service users, the right decision when choosing suitable Cloud services can help them improve user satisfaction and trading revenues. This work aims to address the issue of uncertainty in service requests, service descriptions, user and expert preferences, as well as evaluation criteria in a MCDM-based service selection procedure. A hybrid fuzzy framework for Cloud service selection is proposed, addressing the challenge using three approaches: a fuzzy-ontology-based approach for function matching and service filtering, a fuzzy AHP technique for informed criterion weighting, and, a fuzzy TOPSIS approach for service ranking.

Finding relevant Web services and composing them into value-added applications is becoming increasingly important in cloud and service based marketplaces. The key problem with current approaches to finding relevant Web services is that most of them only provide searches over a discrete set of features using exact keyword matching. We demonstrate in this paper that by utilizing well known indexing scheme such as inverted file and R-tree indexes over Web services attributes, the Earth Mover's Distance (EMD) algorithm can be used efficiently to find partial matches between a query and a database of Web services.

With an ever-increasing number of Web services being available, finding desired Web service is crucial for service users. Current keyword search and most existing approaches are inefficient in two main aspects: poor scalability and lack of semantics. Firstly, users are overwhelmed by the huge number of irrelevant services returned. Secondly, the intentions of users and the semantics in Web services are ignored. Inspired by the success of the divide and conquer approach used to handle the complex information decomposition, we use a novel approach to partition a large set of search results into a set of smaller groups by employing a clustering approach. Then we utilize singular value decomposition (SVD) to capture the main semantics hidden behind the words in a query and the descriptions in the services, so that service matching can be carried out at the concept level. We report here on the preliminary experimental evaluation that shows improvements overall precision.

Service discovery is one of challenging issues in Service-Oriented computing. Currently, most of the existing service discovering and matching approaches are based on keywords-based strategy. However, this method is inefficient and time-consuming. In this paper, we present a novel approach for discovering web services. Based on the current dominating mechanisms of discovering and describing Web Services with UDDI and WSDL, the proposed approach utilizes Probabilistic Latent Semantic Analysis (PLSA) to capture semantic concepts hidden behind words in the query and advertisements in services so that services matching is expected to carry out at concept level. We also present related algorithms and preliminary experiments to evaluate the effectiveness of our approach.