Current data-driven AI approaches in Network Intrusion Detection System (NIDS) face challenges related to high resource consumption, high computational demands, and limited interpretability. Moreover, they often struggle to detect unknown and rapidly evolving cyber threats. This survey explores the integration of Neurosymbolic AI (NeSy AI) into NIDS, combining the data-driven capabilities of Deep Learning (DL) with the structured reasoning of symbolic AI to address emerging cybersecurity threats. The integration of NeSy AI into NIDS demonstrates significant improvements in both the detection and interpretation of complex network threats by exploiting the advanced pattern recognition typical of neural processing and the interpretive capabilities of symbolic reasoning. In this survey, we categorise the analysed NeSy AI approaches applied to NIDS into logic-based and graph-based representations. Logic-based approaches emphasise symbolic reasoning and rule-based inference. On the other hand, graph-based representations capture the relational and structural aspects of network traffic. We examine various NeSy systems applied to NIDS, highlighting their potential and main challenges. Furthermore, we discuss the most relevant issues in the field of NIDS and the contribution NeSy can offer. We present a comparison between the main XAI techniques applied to NIDS in the literature and the increased explainability offered by NeSy systems.

Ever-changing cyber threats require strong and flexible network security solutions. This paper suggests a new method to improve the performance of detecting both known and unknown attacks using a neuro-symbolic artificial intelligence (NSAI) network intrusion detection system (NIDS). Deep neural networks (DNN) learn complex network data patterns, which create a detailed overview of cyber-attack characteristics. Symbolic logic integration into the DNN allows for model training guidance by applying penalties when the DNN fails to differentiate between malicious and benign network traffic. This improves our model’s adaptability to new attacks and overcomes traditional signature-based NIDS limitations. By testing our NSAI NIDS on a large cyber dataset that includes novel attack scenarios, we show that it delivers an improvement in how accurately it detects attacks compared to traditional DNN methods. While our system maintains its high accuracy in recognizing known attacks, it outperforms conventional NIDS in discovering unknown attacks. This work improves cybersecurity by introducing a new way to detect both known and unknown network intrusions by combining DNNs with symbolic logic.

Task offloading is a powerful tool in Mobile Edge Computing (MEC). However, in many practical scenarios, the number of required processing cycles of a task is unknown beforehand and only known until its completion. This poses a serious challenge in making offloading decisions as the number of processing cycles is a key parameter to determine whether a task’s deadline can be met. To cope with such processing uncertainty, we formulate a Chance-Constrained Program (CCP) that offers probabilistic guarantees to task deadlines. The goal is to minimize energy consumption for the users while meeting the probabilistic task deadlines. We assume that only the means and variances of the random processing cycles are available, without any knowledge of distribution functions. We employ a powerful tool called Exact Conic Reformulation (ECR) that reformulates probabilistic deadline constraints into deterministic ones. Subsequently, we design an online solution called EPD (Energy-minimized solution with Probabilistic Deadline guarantee) for periodic scheduling and schedule updates during run-time. We show that EPD can address the processing uncertainty with probabilistic deadline guarantees while minimizing the users’ energy consumption.

Degree-of-freedom (DoF)-based models have been proven to be highly successful in modeling and analysis of MIMO systems. Among existing DoF-based models, the number of DoFs used for interference cancellation (IC) is solely based on the number of interfering data streams. However, from both experimental and simulation results, we find that signal strengths of an interference link vary significantly in different directions in the eigenspace. In this paper, we exploit the difference in interference signal strengths in the eigenspace and perform IC with DoFs only on those directions with strong signals. To differentiate interference signal strengths on an interference link, we introduce a novel concept called “effective rank threshold.” Based on this threshold, DoFs are consumed only to cancel strong interferences in the eigenspace while weak interferences are treated as noise in throughput calculation. To better understand the benefits of this approach, we study a fundamental trade-off between network throughput and effective rank threshold for an MU-MIMO network. Our simulation results show that network throughput under optimal rank threshold is significantly higher than that under existing DoF IC models. To ensure the new DoF IC model is feasible at PHY layer, we propose an algorithm to set the weights for all nodes that can offer our desired DoF allocation.

Underlay coexistence is a spectrum efficient mechanism to roll out 5G picocells within a macrocell on the same spectrum. Due to a lack of cooperation between the primary users (PUs) in the macrocell and secondary users (SUs) in the picocells, it is impossible to have complete knowledge of channel conditions between them. Under such a circumstance, chance-constrained programming (CCP) has been shown to be an ideal optimization tool to address such a channel uncertainty. However, solutions to CCP are computationally intensive and cannot meet 5G’s timing requirement (125$\mu s$). To address this problem, we propose a novel scheduler called GPU-based Underlay Coexistence (GUC) with the goal of finding an approximate solution to CCP in real-time. The essence of GUC is to decompose the original optimization problem into a large number of small subproblems that are suitable for parallel computation on GPU platforms. By selecting a subset of promising subproblems and solving them in parallel with fast algorithms, we are able to leverage GPU parallel computing and develop a real-time solution. Through extensive experiments, we show that GUC meets the 125$\mu s$requirement while achieving 90% optimality on average.

Graph Neural Networks (GNNs) have shown superior performance in analyzing attributed networks in various web-based applications such as social recommendation and web search. Nevertheless, in high-stake decision-making scenarios such as online fraud detection, there is an increasing societal concern that GNNs could make discriminatory decisions towards certain demographic groups. Despite recent explorations on fair GNNs, these works are tailored for a specific GNN model. However, myriads of GNN variants have been proposed for different applications, and it is costly to fine-tune existing debiasing algorithms for each specific GNN architecture. Different from existing works that debias GNN models, we aim to debias the input attributed network to achieve fairer GNNs through feeding GNNs with less biased data. Specifically, we propose novel definitions and metrics to measure the bias in an attributed network, which leads to the optimization objective to mitigate bias. We then develop a framework EDITS to mitigate the bias in attributed networks while maintaining the performance of GNNs in downstream tasks. EDITS works in a model-agnostic manner, i.e., it is independent of any specific GNN. Experiments demonstrate the validity of the proposed bias metrics and the superiority of EDITS on both bias mitigation and utility maintenance. Open-source implementation: https://github.com/yushundong/EDITS.

We study the problem of channel uncertainty on wireless transmissions from different users with mutual interference. Specifically, the channel gains from the transmitters to the receivers are available only through their mean and covariance rather than complete distributions. Our goal is to maximize the energy efficiency among all transmitter-receiver pairs while guaranteeing their capacity requirements. For this problem, we employ chance-constrained programming (CCP), which allows occasional violation of target capacity threshold as long as the probability of such violation is below a small tolerable constant (risk level). We propose a solution based on a novel reformulation technique that converts the original CCP into a deterministic optimization problem without relaxation errors. Then the deterministic optimization problem is approximated into a Geometric Program (GP) based on tight polynomial approximations, which can be solved optimally. We prove that our proposed solution achieves near-optimal performance with polynomial time complexity.

We propose a novel Evolutionary Algorithm (EA) based on the Differential Evolution algorithm for solving global numerical optimization problem in real-valued continuous parameter space. The proposed MadDE algorithm leverages the power of the multiple adaptation strategy with respect to the control parameters and search mechanisms, and is tested on the benchmark functions taken from the CEC 2021 special session & competition on single-objective bound-constrained optimization. Experimental results indicate that MadDE is able to achieve superior performance on global numerical optimization problems when compared against state-of-the-art real-parameter optimizers. We also provide a hyperparameter optimization algorithm SUBHO for improving the search performance of any EA by finding an optimal set of control parameters, and demonstrate its efficacy in enhancing MadDE's performance on the same benchmark. The source code of our implementation is publicly available at https://github.com/subhodipbiswas/MadDE.

Graph Neural Networks have recently become a prevailing paradigm for various high-impact graph analytical problems. Existing efforts can be mainly categorized as spectral-based and spatial-based methods. The major challenge for the former is to find an appropriate graph filter to distill discriminative information from input signals for learning. Recently, myriads of explorations are made to achieve better graph filters, e.g., Graph Convolutional Network (GCN), which leverages Chebyshev polynomial truncation to seek an approximation of graph filters and bridge these two families of methods. Nevertheless, it has been shown in recent studies that GCN and its variants are essentially employing fixed low-pass filters to perform information denoising. Thus their learning capability is rather limited and may over-smooth node representations at deeper layers. To tackle these problems, we develop a novel graph neural network framework AdaGNN with a well-designed adaptive frequency response filter. At its core, AdaGNN leverages a simple but elegant trainable filter that spans across multiple layers to capture the varying importance of different frequency components for node representation learning. The inherent differences among different feature channels are also well captured by the filter. As such, it empowers AdaGNN with stronger expressiveness and naturally alleviates the over-smoothing problem. We empirically validate the effectiveness of the proposed framework on various benchmark datasets. Theoretical analysis is also provided to show the superiority of the proposed AdaGNN. The open-source implementation of AdaGNN can be found here: https://github.com/yushundong/AdaGNN.

Mobile Edge Computing (MEC) has emerged to be an integral component of 5G infrastructure due to its potential to speed up task processing and reduce energy consumption for mobile devices. However, a major technical challenge in making offloading decisions is that the number of required processing cycles of a task is usually unknown in advance. Due to this processing uncertainty, it is difficult to make offloading decisions while providing any guarantee on task deadlines. To address this challenge, we propose EPD---Energy-minimized solution with Probabilistic Deadline guarantee to task offloading problem. The mathematical foundation of EPD is Exact Conic Reformulation (ECR), which is a powerful tool that reformulates a probabilistic constraint for task deadline into a deterministic one. In the absence of distribution knowledge of processing cycles, we use the estimated mean and variance of processing cycles and exploit ECR to the fullest extent in the design of EPD. Simulation results show that EPD successfully guarantees the probabilistic deadlines while minimizing the energy consumption of mobile users, and can achieve significant improvement in energy saving when compared to a state-of-the-art approach.

Hamiltonian Monte Carlo (HMC) is a Markov chain Monte Carlo (MCMC) approach that exhibits favourable exploration properties in high-dimensional models such as neural networks. Unfortunately, HMC has limited use in large-data regimes and little work has explored suitable approaches that aim to preserve the entire Hamiltonian. In our work, we introduce a new symmetric integration scheme for split HMC that does not rely on stochastic gradients. We show that our new formulation is more efficient than previous approaches and is easy to implement with a single GPU. As a result, we are able to perform full HMC over common deep learning architectures using entire data sets. In addition, when we compare with stochastic gradient MCMC, we show that our method achieves better performance in both accuracy and uncertainty quantification. Our approach demonstrates HMC as a feasible option when considering inference schemes for large-scale machine learning problems.

The Age of Information (AoI) is a key metric to measure the freshness of information for IoT applications. Most of the existing analytical models for AoI are overly idealistic and do not capture state-of-the-art transmission technologies such as 5G as well as channel dynamics in both frequency and time domains. In this article, we present Kronos, a real-time 5G-compliant scheduler that minimizes AoI for IoT data collection. Kronos is designed to cope with highly dynamic channel conditions. Its main function is to perform RB allocation and to select the modulation and coding scheme for each source node based on channel conditions, with the objective of minimizing long-term AoI. To meet the stringent real-time requirement for 5G, we develop a GPU-based implementation of Kronos on commercial off-the-shelf Nvidia GPUs. Through extensive experimentation, we show that Kronos can find near-optimal solutions under submillisecond time scale. To the best of our knowledge, this is the first real-time AoI scheduler that is 5G compliant.

We consider an underlay coexistence scenario where secondary users (SUs) must keep their interference to the primary users (PUs) under control. However, the channel gains from the PUs to the SUs are uncertain due to a lack of cooperation between the PUs and the SUs. Under this circumstance, it is preferable to allow the interference threshold of each PU to be violated occasionally as long as such violation stays below a probability. In this article, we employ Chance-Constrained Programming (CCP) to exploit this idea of occasional interference threshold violation. We assume the uncertain channel gains are only known by their mean and covariance. These quantities are slow-changing and easy to estimate. Our main contribution is to introduce a novel and powerful mathematical tool called Exact Conic Reformulation (ECR), which reformulates the intractable chance constraints into tractable convex constraints. Further, ECR guarantees an equivalent reformulation from linear chance constraints into deterministic conic constraints without the limitations associated with Bernstein Approximation, on which our research community has been fixated on for years. Through extensive simulations, we show that our proposed solution offers a significant improvement over existing approaches in terms of performance and ability to handle channel correlations (where Bernstein Approximation is no longer applicable).

In this paper, we present a general framework for distilling expectations with respect to the Bayesian posterior distribution of a deep neural network classifier, extending prior work on the Bayesian Dark Knowledge framework. The proposed framework takes as input "teacher" and "student" model architectures and a general posterior expectation of interest. The distillation method performs an online compression of the selected posterior expectation using iteratively generated Monte Carlo samples. We focus on the posterior predictive distribution and expected entropy as distillation targets. We investigate several aspects of this framework including the impact of uncertainty and the choice of student model architecture. We study methods for student model architecture search from a speed-storage-accuracy perspective and evaluate down-stream tasks leveraging entropy distillation including uncertainty ranking and out-of-distribution detection.

Underlay coexistence is an effective mechanism to improve spectrum effïciency by having picocells coexist with macrocell on the same spectrum. Due to a lack of cooperation between the primary users (PUs) in the macrocell and secondary users (SUs) in the picocell, it is impossible to have complete knowledge of channel gains between them. Under such circumstance, chance-constrained programming (CCP) is shown to be the ideal optimization tool to address such uncertainty. However, solutions to CCP are computationally intensive and cannot meet 5G's timing requirement. To address this problem, we propose a novel scheduler called GUC (stands for GPU-based Underlay Coexistence) to find an approximate solution to CCP in real-time. The essence of GUC is to decompose the original optimization problem into a large number of small problems that are suitable for parallel computation on GPU platforms. Through extensive experiments, we show that GUC reduces the scheduling computation time by at least 10,000 times comparing to commercial solvers (on CPU) while achieving an average of 90% optimality.

Age of information (AoI) is a powerful new metric to quantify the freshness of information and has gained increasing popularity in IoT applications. Existing models on AoI remain primitive and do not consider state-of-the-art transmission technologies such as 5G. They also fail to consider the impact of dynamic channel conditions. In this paper, we present Kronos, a 5G-compliant AoI scheduling algorithm that can cope with highly dynamic channel conditions. Kronos is capable of performing RB allocation and selecting MCS for each source node based on channel conditions, with the objective of minimizing long-term AoI. To meet the stringent real-time requirement for 5G, we propose a GPU-based implementation of Kronos on low-cost offthe-shelf GPUs. Through simulations and experiments, we show that Kronos can find near-optimal AoI scheduling solutions in sub-millisecond time scale. To the best of our knowledge, this is the first 5G-compliant real-time AoI scheduler that can cope with dynamic channel conditions.

Degree-of-Freedom (DoF) based models have been widely used to study MIMO networks. To cancel interference, the number of DoFs used in the state-of-the-art DoF models is solely based on the number of interfering data streams. However, by decomposing an interference into the eigenspace, we find that signal strengths varies significantly in different directions for the same interference link. In this paper, we exploited the difference in interference signal strength in the eigenspace and differentiate strong and weak interference signals via their singular values. By introducing a concept of effective rank threshold, we propose to use DoFs only to cancel strong interference in the eigenspace based on this threshold while treating weak interference signals as noise in throughput calculation. We explore a fundamental tradeoff between network throughput and effective rank threshold. Using simulation results on MU-MIMO networks, we show that network throughput under optimal rank threshold setting is significantly higher than that under existing DoF IC models. To ensure feasibility at the PHY layer, we present an algorithm that can find Tx and Rx weights at each node that can offer our desired DoF allocation.

In underlay coexistence, secondary users (SUs) attempt to keep their interference to the primary users (PUs) under a threshold. Due to the absence of cooperation from the PUs, there exists much uncertainty at the SUs in terms of channel state information (CSI). An effective approach to cope such uncertainty is to introduce occasional interference threshold violation by the SUs, as long as such occasional violation can be tolerated by the PUs. This paper exploits this idea through a chance constrained programming (CCP) formulation, where the knowledge of uncertain CSI is limited to only the first and second order statistics rather than its complete distribution information. Our main contribution is the introduction of a novel and powerful technique, called Exact Conic Reformulation (ECR), to reformulate the intractable chance constraints. ECR guarantees an equivalent reformulation for linear chance constraints into deterministic conic constraints and does not suffer from the limitations associated with the state-of-the-art approach -- Bernstein Approximation. Simulation results confirm that ECR offers significant performance improvement over Bernstein Approximation in uncorrelated channels and a competitive solution in correlated channels (where Bernstein Approximation is no longer applicable).

We propose a novel confidence metric, namely, attribution-based confidence (ABC) for deep neural networks (DNNs). ABC metric characterizes whether the output of a DNN on an input can be trusted. DNNs are known to be brittle on inputs outside the training distribution and are, hence, susceptible to adversarial attacks. This fragility is compounded by a lack of effectively computable measures of model confidence that correlate well with the accuracy of DNNs. These factors have impeded the adoption of DNNs in high-assurance systems. The proposed ABC metric addresses these challenges. It does not require access to the training data, the use of ensembles, or the need to train a calibration model on a held-out validation set. Hence, the new metric is usable even when only a trained model is available for inference. We mathematically motivate the proposed metric and evaluate its effectiveness with two sets of experiments. First, we study the change in accuracy and the associated confidence over out-of-distribution inputs. Second, we consider several digital and physically realizable attacks such as FGSM, CW, DeepFool, PGD, and adversarial patch generation methods. The ABC metric is low on out-of-distribution data and adversarial examples, where the accuracy of the model is also low. These experiments demonstrate the effectiveness of the ABC metric to make DNNs more trustworthy and resilient.

There have been some rapid advances on the design of full duplex (FD) transceivers in recent years. Although the benefits of FD have been studied for single-hop wireless communications, its potential on throughput performance in a multi-hop wireless network remains unclear. As for multi-hop networks, a fundamental problem is to compute the achievable end-to-end throughput for one or multiple communication sessions. The goal of this paper is to offer some fundamental understanding on end-to-end throughput performance limits of FD in a multi-hop wireless network. We show that through a rigorous mathematical formulation, we can cast the multi-hop throughput performance problem into a formal optimization problem. Through numerical results, we show that in many cases, the end-to-end session throughput in a FD network can exceed 2x of that in a half duplex (HD) network. Our finding can be explained by the much larger design space for scheduling that is offered by removing HD constraints in throughput maximization problem. The results in this paper offer some new understandings on the potential benefits of FD for end-to-end session throughput in a multi-hop wireless network.

Recent advances in MIMO degree-of-freedom (DoF)models allow us to study MIMO in a multi-hop network environment. On the other hand, successive interference cancellation (SIC) is a powerful physical layer technique used in multi-user detection. Based on the strengths and weaknesses of MIMO DoF and SIC, we propose a marriage between these two techniques so that DoF-based interference cancellation (IC) and SIC can help each other as follows: (i) SIC is exploited to decode multiple received signals to conserve DoF resources in IC, and (ii) DoFIC resolves the potential SINR barrier that SIC may encounter. In this paper, we develop the necessary mathematical models to realize the two ideas in a multi-hop wireless network. Together with scheduling and routing constraints, we develop a cross-layer optimization framework with joint DoF IC and SIC. By applying the framework on a throughput maximization problem, we find that SIC and DoF IC can indeed offer significant performance improvement by addressing each other's limitation.