• 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.

Device authentication in Low Earth Orbit (LEO) satellite-based Internet of Things (IoT) networks is critical for enabling secure and reliable communication between remote IoT devices and satellites. It prevents unauthorized access and security breaches. State-of-the-art authentication methods for terrestrial networks, such as Authentication and Key Management for Applications (AKMA), are inadequate when directly applied to such networks because IoT devices have constrained communication and computational capabilities. Further, the satellite environment is highly dynamic, with frequent handovers and variable latency, leading to vulnerabilities like man-in-the-middle (MITM) and spoofing attacks. To address these challenges, we propose a modified AKMA framework for decentralized and continuous authentication in LEO satellite-based IoT networks. Our proposed modification utilizes local key refreshment for seed generation, seed update, and seed refreshment in a decentralized manner, enabling tailored transmission patterns for IoT devices. This reduces the need for repeated authentication attempts with satellites and effectively mitigates handoff-associated threats. We examine the authentication performance of the system in the presence of an illegitimate Unmanned Aerial Vehicle (UAV) above the legitimate IoT devices. Our results through simulations and emulation show improvement in the authentication rate of legitimate IoT devices and a reduction in the misdetection rate of illegitimate UAVs compared to state-of-the-art physical channel-based authentication schemes. Our proposed modified AKMA enables its application in LEO satellite-based IoT networks.

Recent advancements have sparked a growing interest in integrating spatiotemporal analysis with large-scale language models. However, spatiotemporal data often contains sensitive information, making it unsuitable for open third-party access. To address this challenge, we propose a Graph-GAN-based model for generating privacy-protected spatiotemporal data. Our approach incorporates spatial and temporal attention blocks in the discriminator and a spatiotemporal deconvolution structure in the generator. These enhancements enable efficient training under Gaussian noise to achieve differential privacy. Extensive experiments conducted on three real-world spatiotemporal datasets validate the efficacy of our model. Our method provides a privacy guarantee while maintaining the data utility. The prediction model trained on our generated data maintains a competitive performance compared to the model trained on the original data.

Protecting sensitive information on data streams is a pivotal challenge for modern systems. Current approaches to providing privacy in data streams can be broadly categorized into two strategies. The first strategy involves transforming the stream into a private sequence of values, enabling the subsequent use of non-private methods of analysis. While effective, this approach incurs high memory costs, often proportional to the size of the database. Alternatively, a compact data structure can be used to provide a private summary of the stream. However, these data structures are limited to predefined queries, restricting their flexibility. To overcome these limitations, we propose a lightweight synthetic data generator, PrivHP, that provides differential privacy guarantees. PrivHP is based on a novel method for the private hierarchical decomposition of the input domain in bounded memory. As the decomposition approximates the cumulative distribution function of the input, it serves as a lightweight structure for synthetic data generation. PrivHP is the first method to provide a principled trade-off between accuracy and space for private hierarchical decompositions. It achieves this by balancing hierarchy depth, noise addition, and selective pruning of low-frequency subdomains while preserving high-frequency ones, all identified in a privacy-preserving manner. To ensure memory efficiency, we employ private sketches to estimate subdomain frequencies without accessing the entire dataset. Central to our approach is the introduction of a pruning parameter k , which enables an almost smooth interpolation between space usage and utility, and a measure of skew tail k , which is a vector of subdomain frequencies containing all but the largest k coordinates. PrivHP processes a dataset X using M = O (k log 2 | X |)) space and, on input domain Ω = [0,1] d , while maintaining ε-differential privacy, produces a synthetic data generator that is at distance O ( M (1-1/d) /ε n + ||tail k ( X )|| 1 /M 1/d n ) from the empirical distribution in the expected Wasserstein metric. Compared to the state-of-the-art, PMM, which achieves accuracy O ((ε n) -1/d ) with memory O (ε n), our method introduces an additional approximation error term of O (||tail k ( X )|| 1 /(M 1/d n)), but operates in significantly reduced space. Additionally, we provide interpretable utility bounds that account for all error sources, including those introduced by the fixed hierarchy depth, privacy noise, hierarchy pruning, and frequency approximations.

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.

Large machine-learning training datasets can be distilled into small collections of informative synthetic data samples. These synthetic sets support efficient model learning and reduce the communication cost of data sharing. Thus, high-fidelity distilled data can support the efficient deployment of machine learning applications in distributed network environments. A naive way to construct a synthetic set in a distributed environment is to allow each client to perform local data distillation and to merge local distillations at a central server. However, the quality of the resulting set is impaired by heterogeneity in the distributions of the local data held by clients. To overcome this challenge, we introduce the first collaborative data distillation technique, called CollabDM, which captures the global distribution of the data and requires only a single round of communication between client and server. Our method outperforms the state-of-the-art one-shot learning method on skewed data in distributed learning environments. We also show the promising practical benefits of our method when applied to attack detection in 5G networks.

Physical-layer authentication is a popular alternative to the conventional key-based authentication for Internet of Things (IoT) devices due to their limited computational capacity and battery power. However, this approach has limitations due to poor robustness under channel fluctuations, reconciliation overhead, and no clear safeguard distance to ensure the secrecy of the generated authentication keys. In this regard, we propose a novel, secure, and lightweight continuous authentication scheme for IoT device authentication. Our scheme utilizes the inherent properties of the IoT devices’ transmission model as its source for seed generation and device authentication. Specifically, our proposed scheme provides continuous authentication by checking the access time slots and spreading sequences of the IoT devices instead of repeatedly generating and verifying shared keys. Due to this, access to a coherent key is not required in our proposed scheme, resulting in the concealment of the seed information from attackers. Our proposed authentication scheme for IoT devices demonstrates improved performance compared to the benchmark schemes relying on physical channels. Our empirical results find a near threefold decrease in the misdetection rate of illegitimate devices and close to zero false alarm rate in various system settings with varied numbers of active devices up to 200 and signal-to-noise ratio from 0 to 25 dB. Our proposed authentication scheme also has a lower computational complexity of at least half the computational cost of the benchmark schemes based on support vector machine and binary hypothesis testing in our studies. This further corroborates the practicality of our scheme for IoT deployments.

Moving Target Defenses (MTD) are proactive security countermeasures that change the attack surface in a system in ways that make it harder for attackers to succeed. These techniques have been shown to be effective, and their application in software-defined networking (SDN) against simple automated attacks is growing in popularity. However, with the increased knowledge of and ease of access to Artificial Intelligence (AI) techniques, AI is starting to be used to enhance cyber attacks, which are becoming increasingly complex. Hence, the evaluation of MTDs against simple automated attacks is no longer enough to demonstrate their effectiveness in increasing system security.

Federated learning (FL) and split learning (SL) are two popular distributed machine learning approaches. Both follow a model-to-data scenario; clients train and test machine learning models without sharing raw data. SL provides better model privacy than FL due to the machine learning model architecture split between clients and the server. Moreover, the split model makes SL a better option for resource-constrained environments. However, SL performs slower than FL due to the relay-based training across multiple clients. In this regard, this paper presents a novel approach, named splitfed learning (SFL), that amalgamates the two approaches eliminating their inherent drawbacks, along with a refined architectural configuration incorporating differential privacy and PixelDP to enhance data privacy and model robustness. Our analysis and empirical results demonstrate that (pure) SFL provides similar test accuracy and communication efficiency as SL while significantly decreasing its computation time per global epoch than in SL for multiple clients. Furthermore, as in SL, its communication efficiency over FL improves with the number of clients. Besides, the performance of SFL with privacy and robustness measures is further evaluated under extended experimental settings.

The large transformer-based language models demonstrate excellent performance in natural language processing. By considering the transferability of the knowledge gained by these models in one domain to other related domains, and the closeness of natural languages to high-level programming languages, such as C/C++, this work studies how to leverage (large) transformer-based language models in detecting software vulnerabilities and how good are these models for vulnerability detection tasks. In this regard, firstly, we present a systematic (cohesive) framework that details source code translation, model preparation, and inference. Then, we perform an empirical analysis of software vulnerability datasets of C/C++ source codes having multiple vulnerabilities corresponding to the library function call, pointer usage, array usage, and arithmetic expression. Our empirical results demonstrate the good performance of the language models in vulnerability detection. Moreover, these language models have better performance metrics, such as F1-score, than the contemporary models, namely bidirectional long short term memory and bidirectional gated recurrent unit. Experimenting with the language models is always challenging due to the requirement of computing resources, platforms, libraries, and dependencies. Thus, this paper also analyses the popular platforms to efficiently fine-tune these models and present recommendations while choosing the platforms for our framework.

Training machine learning algorithms on sensitive, illegal to possess, and psychologically harmful data is challenging because researchers have to do training without handling the data. Moreover, the nature of the data imposes strict control, monitoring, and examination of all the activities involved, including communication, execution, and release of algorithms, datasets, outputs, and results. In this regard, this work proposes a new multi-zoned framework called MaLFraDA. MaLFraDA has soft air gaps between its zones to isolate and control communication in and out of the framework. Besides, it includes (i) a vetter to investigate and approve incoming model/algorithm, and outgoing information, (ii) encrypted data vaults, and (iii) airlock instances for secure execution/computation. MaLFraDA, with an extension, runs popular distributed machine learning algorithms such as federated and split learning using multiple data custodians.

With rising temperatures and events contributing to climate change, the world is facing extreme weather patterns. Recently, Australia was hit hard by bushfires, the most devastating fires ever faced by the country. The economic damage reported was nearly one billion Australian dollars and an estimated three billion native animals were killed or adversely affected. Given the extent and intensity of this damage, researchers are seeking effective solutions to enable the prediction of fire before it starts to increase the time available for firefighters to protect lives and assets and prepare to mitigate the fires. This motivated us to investigate an approach to address this critical problem. In this article, we propose a machine learning (ML)-based approach that detects anomalies in spatiotemporal measurements of environmental parameters (e.g., temperature, relative humidity, etc.). In the proposed approach, an ML-based model learns the normal spatiotemporal behavior of the environmental data (collected over a period of one year). This is carried out during a one-time training phase. Then, during the detection phase, any spatiotemporal pattern in the real-time data (received from the field sensors) that is different than the normal pattern will be identified by the model as anomaly which indicates a possible bushfire situation. Following this, we propose a supplementary classification model based on Moran’s I index to ensure that the detected anomalies are not due to either a sensor failure or a security attack (which are common in Internet of Things). We developed three different ML models for performance evaluation and comparison and used the Forest Fire data set to train them. The results of our experiments confirm the effectiveness of the proposed approach in the early detection of fire symptoms.

Federated learning (FL) and split learning (SL) are state-of-the-art distributed machine learning techniques to enable machine learning training without accessing raw data on clients or end devices. However, their comparative training performance under real-world resource-restricted Internet of Things (IoT) device settings remains barely studied. This work provides empirical comparisons of FL and SL in real-world IoT settings regarding (i) learning performance with heterogeneous data distributions and (ii) on-device execution overhead. Our analyses in this work demonstrate that the learning performance of SL is better than FL under an imbalanced data distribution but worse than FL under an extreme non-IID data distribution. Recently, FL and SL are combined to form splitfed learning (SFL) to leverage each of their benefits (e.g., parallel training of FL and lightweight on-device computation requirement of SL). Our work considers FL, SL, and SFL, and mounts them on Raspberry Pi devices to evaluate their performance, including training time, communication overhead, power consumption, and memory usage with resource-restricted IoT devices. Besides evaluations, we apply two optimizations. First, we generalize SFL by carefully examining the possibility of a hybrid type of model training at the server-side. The generalized SFL merges sequential (dependent) and parallel (independent) processes of model training and thus is beneficial to a system with a large scale of IoT devices, specifically at the server-side operations. Second, we propose pragmatic techniques to substantially reduce the communication overhead by up to four times for the SL and (generalized) SFL.

A new collaborative learning, called split learning, was recently introduced, aiming to protect user data privacy without revealing raw input data to a server. It collaboratively runs a deep neural network model where the model is split into two parts, one for the client and the other for the server. Therefore, the server has no direct access to raw data processed at the client. Until now, the split learning is believed to be a promising approach to protect the client's raw data; for example, the client's data was protected in healthcare image applications using 2D convolutional neural network (CNN) models. However, it is still unclear whether the split learning can be applied to other deep learning models, in particular, 1D CNN.

Internet of Medical Things (IoMT) are getting popular in the smart healthcare domain. These devices are resource-constrained and are vulnerable to attack. As the IoMTs are connected to the healthcare network infrastructure, it becomes the primary target of the adversary due to weak security and privacy measures. In this regard, this paper proposes a security architecture for smart healthcare network infrastructures. The architecture uses various security components or services that are developed and deployed as virtual network functions. This makes the security architecture ready for future network frameworks such as OpenMANO. Besides, in this security architecture, only authenticated and trusted IoMTs serve the patients along with an encryption-based communication protocol, thus creating a secure, privacy-preserving and trusted healthcare network infrastructure.

Federated learning (FL) and split neural networks (SplitNN) are state-of-art distributed machine learning techniques to enable machine learning without directly accessing raw data on clients or end devices. In theory, such distributed machine learning techniques have great potential in distributed applications, in which data are typically generated and collected at the client-side while the collected data should be processed by the application deployed at the server-side. However, there is still a significant gap in evaluating the performance of those techniques concerning their practicality in the Internet of Things (IoT)-enabled distributed systems constituted by resource-constrained devices. This work is the first attempt to provide empirical comparisons of FL and SplitNN in real-world IoT settings in terms of learning performance and device implementation overhead. We consider a variety of datasets, different model architectures, multiple clients, and various performance metrics. For the learning performance (i.e., model accuracy and convergence time), we empirically evaluate both FL and SplitNN under different types of data distributions such as imbalanced and non-independent and identically distributed (non-IID) data. We show that the learning performance of SplitNN is better than FL under an imbalanced data distribution but worse than FL under an extreme non-IID data distribution. For implementation overhead, we mount both FL and SplitNN on Raspberry Pi devices and comprehensively evaluate their overhead, including training time, communication overhead, power consumption, and memory usage. Our key observations are that under the IoT scenario where the communication traffic is the primary concern, FL appears to perform better over SplitNN because FL has a significantly lower communication overhead compared with SplitNN. However, our experimental results also demonstrate that neither FL or SplitNN can be applied to a heavy model, e.g., with several million parameters, on resource-constrained IoT devices because its training cost would be too expensive for such devices. Source code is released and available: https://github.com/Minki-Kim95/Federated-Learning-and-Split-Learning-with-raspberry-pi.

We provide a correction to[1]in response to an error reported by Vaddi and Rajan[2]. To this effect, we add one extra condition for the definition of an$\mathsf {IC}$structure on page 3696.

We consider a graphical approach to index coding. As cycles have been shown to provide coding gain, cycles and cliques (a specific type of overlapping cycles) have been exploited in an existing literature. In this paper, we define a more general form of overlapping cycles, called the interlinked-cycle (IC) structure, that generalizes cycles and cliques. We propose a scheme, called the interlinked-cycle-cover (ICC) scheme, that leverages IC structures in digraphs to construct scalar linear index codes. We characterize a class of infinitely many digraphs where our proposed scheme is optimal over all linear and nonlinear index codes. Consequently, for this class of digraphs, we indirectly prove that scalar linear index codes are optimal. Furthermore, we show that the ICC scheme can outperform all the existing graph-based schemes (including partial-clique-cover and fractional-local-chromatic number schemes), and a random coding scheme (namely, composite coding) for certain graphs.

We study the index coding problem in the unicast message setting, i.e., where each message is requested by one unique receiver. This problem can be modeled by a directed graph. We propose a new scheme called interlinked cycle cover, which exploits interlinked cycles in the directed graph, for designing index codes. This new scheme generalizes the existing clique cover and cycle cover schemes. We prove that for a class of infinitely many digraphs with messages of any length, interlinked cycle cover provides an optimal index code. Furthermore, the index code is linear with linear time encoding complexity.