arXiv subjects

Mahmood Ahmadi

Publications and source records attributed to Mahmood Ahmadi.

At least 19 recordsLinked to original sources

arXiv2026

Enhancing Error Detection Performance through Parallel CRC Computation on Multi-Core Architectures

Cyclic Redundancy Check (CRC) remains one of the most widely used error-detection mechanisms in communication, storage, and embedded systems. However, conventional software CRC implementations suffer from inherent sequential dependencies that limit efficient utilization of modern multi-core processors. This paper presents a generalized software-based parallel CRC framework for multi-core architectures using POSIX threads. The proposed framework supports multiple CRC variants, including CRC-8, CRC-16, CRC-32, CRC-64, and CRC-128, within a unified implementation model. To preserve correctness during parallel execution, the framework employs a GF(2)-based CRC combination mechanism rather than naive XOR aggregation. The combine stage is formulated using polynomial arithmetic and matrix-based shifting operations over GF(2), ensuring equivalence between parallel and serial CRC computation. The proposed method was evaluated using multiple workload sizes and thread configurations. Experimental analysis includes execution time, throughput, latency, scalability behavior, and energy estimation under varying thread counts. Results indicate that parallel execution significantly improves performance for large datasets, achieving approximately 3-4x speedup on the evaluated platform while preserving exact CRC correctness. Comparative discussion with representative CRC optimization approaches, including lookup-table methods, slicing-by-8, SIMD/vectorized CRC, and hardware-assisted CRC techniques, is also provided to position the proposed framework within the broader CRC optimization landscape. Overall, the proposed approach provides a portable and generalized software framework for correctness-preserving parallel CRC acceleration on general-purpose multi-core systems.

cs.DC↗

arXiv2026

A Q-Learning Approach for Dynamic Resource Management in Three-Tier Vehicular Fog Computing

In this paper, a method for predicting the resources required for an intelligent vehicle client using a three-layer vehicular computing architecture is proposed. This method leverages Q-Learning to optimize resource allocation and enhance overall system performance. This approach employs reinforcement learning capabilities to provide a dynamic and adaptive strategy for resource management in a fog computing environment. The key findings of this study indicate that Q-learning can effectively predict the appropriate allocation of resources by learning from past experiences and making informed decisions. Through continuous training and updating of the Q-learning agent, the system can adapt to changing conditions and make resource allocation decisions based on real-time information. The experimental results demonstrate the effectiveness of the proposed method in optimizing resource allocation. The Q-learning agent predicts the optimal values for memory, bandwidth, and processor. These predictions not only minimize resource consumption but also meet the performance requirements of the fog system. Implementations show that this method improves the average task processing time in compared to other methods evaluated in this study

cs.NI↗

arXiv2026

ASA: Adaptive Smart Agent Federated Learning via Device-Aware Clustering for Heterogeneous IoT

Federated learning (FL) has become a promising answer to facilitating privacy-preserving collaborative learning in distributed IoT devices. However, device heterogeneity is a key challenge because IoT networks include devices with very different computational powers, memory availability, and network environments. To this end, we introduce ASA (Adaptive Smart Agent). This new framework clusters devices adaptively based on real-time resource profiles and adapts customized models suited to every cluster's capability. ASA capitalizes on an intelligent agent layer that examines CPU power, available memory, and network environment to categorize devices into three levels: high-performance, mid-tier, and low-capability. Each level is provided with a model tuned to its computational power to ensure inclusive engagement across the network. Experimental evaluation on two benchmark datasets, MNIST and CIFAR-10, proves that ASA decreases communication burden by 43% to 50%, improves resource utilization by 43%, and achieves final model accuracies of 98.89% on MNIST and 85.30% on CIFAR-10. These results highlight ASA's efficacy in enhancing efficiency, scalability, and fairness in heterogeneous FL environments, rendering it a suitable answer for real-world IoT apps.

cs.NI↗

arXiv2026

Adaptive Meta-Aggregation Federated Learning for Intrusion Detection in Heterogeneous Internet of Things

The rapid proliferation of the Internet of Things (IoT) has brought remarkable advancements to industries by enabling interconnected systems and intelligent automation. However, this exponential growth has also introduced significant security vulnerabilities, making IoT networks increasingly targets for sophisticated cyberattacks. The heterogeneity of IoT devices poses critical challenges for traditional intrusion detection systems. To address these challenges, this paper proposes an innovative method called Adaptive Meta-Aggregation Federated Learning (AMAFed), designed to enhance intrusion detection in heterogeneous IoT networks. By employing a dynamic weighting mechanism using meta-learning, AMAFed assigns adaptive importance to local models based on their data quality and contributions, enabling personalized yet collaborative learning across devices. The proposed method was evaluated on three benchmark IoT datasets: ToN-IoT, N-BaIoT, and BoT-IoT, representing diverse real-world scenarios. Experimental results demonstrate that AMAFed achieves detection accuracy up to 99.8% on ToN-IoT, with F1-scores exceeding 98% across all datasets. On the N-BaIoT dataset, it reaches 99.88% accuracy, and on BoT-IoT, it achieves 98.12% accuracy, consistently outperforming state-of-the-art approaches.

cs.NI↗

arXiv2026

Lightweight Cluster-Based Federated Learning for Intrusion Detection in Heterogeneous IoT Networks

The rise of heterogeneous Internet of Things (IoT) devices has raised security concerns due to their vulnerability to cyberattacks. Intrusion Detection Systems (IDS) are crucial in addressing these threats. Federated Learning (FL) offers a privacy-preserving solution, but IoT heterogeneity and limited computational resources cause increased latency and reduced performance. This paper introduces a novel approach Cluster-based federated intrusion detection with lightweight networks for heterogeneous IoT designed to address these limitations. The proposed framework utilizes a hierarchical IoT architecture that encompasses edge, fog, and cloud layers. Intrusion detection clients operate at the fog layer, leveraging federated learning to enhance data privacy and distributed processing efficiency. To enhance efficiency, the method employs the lightweight MobileNet model alongside a hybrid loss function that integrates Gumbel-SoftMax and SoftMax, optimizing resource consumption while maintaining high detection accuracy. A key feature of this approach is clustering IoT devices based on hardware similarities, enabling more efficient model training and aggregation tailored to each cluster's computational capacity. This strategy not only simplifies the complexity of managing heterogeneous data and devices but also improves scalability and overall system performance. To validate the effectiveness of the proposed method, extensive experiments were conducted using the ToN-IoT and CICDDoS2019 datasets. Results demonstrate that the proposed approach reduces end-to-end training time by 2.47x compared to traditional FL methods, achieves 2.16x lower testing latency, and maintains exceptionally high detection accuracy of 99.22% and 99.02% on the ToN-IoT and CICDDoS2019 datasets, respectively.

cs.NI↗

arXiv2026

CF-HFC:Calibrated Federated based Hardware-aware Fuzzy Clustering for Intrusion Detection in Heterogeneous IoTs

The rapid expansion of heterogeneous Internet of Things (IoT) environments has heightened security risks, as resource-constrained devices remain vulnerable to diverse cyberattacks. Federated Learning (FL) has emerged as a privacy-preserving paradigm for collaborative intrusion detection; however, device and data heterogeneity introduce major challenges, including straggler delays, unstable convergence, and unbalanced error rates. This paper presents a Calibrated Federated Learning method with Hardware-aware Fuzzy Clustering (CF-HFC) to enhance intrusion detection performance in heterogeneous IoT networks. The proposed three-tier Edge-Fog-Cloud architecture integrates three complementary components: (1) hardware-aware fuzzy clustering, which organizes clients by computational capacity to mitigate straggler effects; (2) Fuzzy-FedProx aggregation, which stabilizes optimization under non-IID data distributions; and (3) Adaptive Conformal Calibration (ACC), which dynamically adjusts decision thresholds to balance false negative and false positive rates. Extensive experiments on ToN-IoT, BoT-IoT, Edge-IIoTset, and CICDDoS2019 datasets demonstrate that CF-HFC outperforms baseline methods such as FedAvg and FedProx, achieving over 99% detection accuracy, faster convergence, and lower communication latency. Overall, the results verify that CF-HFC effectively mitigates both device- and data-level heterogeneity, compared to existing federated learning approaches, providing accurate and efficient intrusion detection across Heterogeneous IoTs environment.

cs.NI↗

arXiv2025

Hierarchical Reinforcement Learning for Integrated Cloud-Fog-Edge Computing in IoT Systems

The Internet of Things (IoT) is transforming industries by connecting billions of devices to collect, process, and share data. However, the massive data volumes and real-time demands of IoT applications strain traditional cloud computing architectures. This paper explores the complementary roles of cloud, fog, and edge computing in enhancing IoT performance, focusing on their ability to reduce latency, improve scalability, and ensure data privacy. We propose a novel framework, the Hierarchical IoT Processing Architecture (HIPA), which dynamically allocates computational tasks across cloud, fog, and edge layers using machine learning. By synthesizing current research and introducing HIPA, this paper highlights how these paradigms can create efficient, secure, and scalable IoT ecosystems.

cs.NI↗

arXiv2025

Mist-Assisted Federated Learning for Intrusion Detection in Heterogeneous IoT Networks

The rapid growth of the Internet of Things (IoT) offers new opportunities but also expands the attack surface of distributed, resource-limited devices. Intrusion detection in such environments is difficult due to data heterogeneity from diverse sensing modalities and the non-IID distribution of samples across clients. Federated Learning (FL) provides a privacy-preserving alternative to centralized training, yet conventional frameworks struggle under these conditions. To address this, we propose a Mist-assisted hierarchical framework for IoT intrusion detection. The architecture spans four layers: (i) Mist, where raw data are abstracted into a unified feature space and lightweight models detect anomalies; (ii) Edge, which applies utility-based client selection; (iii) Fog, where multiple regional aggregators use FedProx to stabilize training; and (iv) Cloud, which consolidates and disseminates global models. Evaluations on the TON-IoT dataset show the framework achieves 98-99% accuracy, PR-AUC> 0.97, and stable convergence under heterogeneous and large-scale settings, while maintaining efficiency and preserving privacy.

cs.NI↗

arXiv2025

Reinforcement Learning for Resource Allocation in Vehicular Multi-Fog Computing

The exponential growth of Internet of Things (IoT) devices, smart vehicles, and latency-sensitive applications has created an urgent demand for efficient distributed computing paradigms. Multi-Fog Computing (MFC), as an extension of fog and edge computing, deploys multiple fog nodes near end users to reduce latency, enhance scalability, and ensure Quality of Service (QoS). However, resource allocation in MFC environments is highly challenging due to dynamic vehicular mobility, heterogeneous resources, and fluctuating workloads. Traditional optimization-based methods often fail to adapt to such dynamics. Reinforcement Learning (RL), as a model-free decision-making framework, enables adaptive task allocation by continuously interacting with the environment. This paper formulates the resource allocation problem in MFC as a Markov Decision Process (MDP) and investigates the application of RL algorithms such as Q-learning, Deep Q-Networks (DQN), and Actor-Critic. We present experimental results demonstrating improvements in latency, workload balance, and task success rate. The contributions and novelty of this study are also discussed, highlighting the role of RL in addressing emerging vehicular computing challenges.

cs.NI↗

arXiv2025

Sentiment Analysis in Twitter Social Network Centered on Cryptocurrencies Using Machine Learning

Cryptocurrency is a digital currency that uses blockchain technology with secure encryption. Due to the decentralization of these currencies, traditional monetary systems and the capital market of each they, can influence a society. Therefore, due to the importance of the issue, the need to understand public opinion and analyze people's opinions in this regard increases. To understand the opinions and views of people about different topics, you can take help from social networks because they are a rich source of opinions. The Twitter social network is one of the main platforms where users discuss various topics, therefore, in the shortest time and with the lowest cost, the opinion of the community can be measured on this social network. Twitter Sentiment Analysis (TSA) is a field that analyzes the sentiment expressed in tweets. Considering that most of TSA's research efforts on cryptocurrencies are focused on English language, the purpose of this paper is to investigate the opinions of Iranian users on the Twitter social network about cryptocurrencies and provide the best model for classifying tweets based on sentiment. In the case of automatic analysis of tweets, managers and officials in the field of economy can gain knowledge from the general public's point of view about this issue and use the information obtained in order to properly manage this phenomenon. For this purpose, in this paper, in order to build emotion classification models, natural language processing techniques such as bag of words (BOW) and FastText for text vectorization and classical machine learning algorithms including KNN, SVM and Adaboost learning methods Deep including LSTM and BERT model were used for classification, and finally BERT linguistic model had the best accuracy with 83.50%.

cs.CL↗

arXiv2024

Grey Wolf-Based Task Scheduling in Vehicular Fog Computing Systems

Vehicular fog computing (VFC) can be considered as an important alternative to address the existing challenges in intelligent transportation systems (ITS). The main purpose of VFC is to perform computational tasks through various vehicles. At present, VFCs include powerful computing resources that bring the computational resources nearer to the requesting devices. This paper presents a new algorithm based on meta-heuristic optimization method for task scheduling problem in VFC. The task scheduling in VFC is formulated as a multi-objective optimization problem, which aims to reduce makespan and monetary cost. The proposed method utilizes the grey wolf optimization (GWO) and assigns the different priorities to static and dynamic fog nodes. Dynamic fog nodes represent the parked or moving vehicles and static fog nodes show the stationary servers. Afterwards, the tasks that require the most processing resources are chosen and allocated to fog nodes. The GWO-based method is extensively evaluated in more details. Furthermore, the effectiveness of various parameters in GWO algorithm is analyzed. We also assess the proposed algorithm on real application and random data. The outcomes of our experiments confirm that, in comparison to previous works, our algorithm is capable of offering the lowest monetary cost.

cs.NI↗

arXiv2024

Deep Reinforcement Learning for Delay-Optimized Task Offloading in Vehicular Fog Computing

The imminent rise of autonomous vehicles (AVs) is revolutionizing the future of transport. The Vehicular Fog Computing (VFC) paradigm has emerged to alleviate the load of compute-intensive and delay-sensitive AV programs via task offloading to nearby vehicles. Effective VFC requires an intelligent and dynamic offloading algorithm. As a result, this paper adapts Deep Reinforcement Learning (DRL) for VFC offloading. First, a simulation environment utilizing realistic hardware and task specifications, in addition to a novel vehicular movement model based on grid-planned cities, is created. Afterward, a DRL-based algorithm is trained and tested on the environment with the goal of minimizing global task delay. The DRL model displays impressive results, outperforming other greedy and conventional methods. The findings further demonstrate the effectiveness of the DRL model in minimizing queue congestion, especially when compared to traditional cloud computing methods that struggle to handle the demands of a large fleet of vehicles. This is corroborated by queuing theory, highlighting the self-scalability of the VFC-based DRL approach.

cs.NI↗

arXiv2024

Machine Learning Approaches for Active Queue Management: A Survey, Taxonomy, and Future Directions

Active Queue Management (AQM), a network-layer congestion control technique endorsed by the Internet Engineering Task Force (IETF), encourages routers to discard packets before the occurrence of buffer overflow. Traditional AQM techniques often employ heuristic approaches that require meticulous parameter adjustments, limiting their real-world applicability. In contrast, Machine Learning (ML) approaches offer highly adaptive, data-driven solutions custom to dynamic network conditions. Consequently, many researchers have adapted ML for AQM throughout the years, resulting in a wide variety of algorithms ranging from predicting congestion via supervised learning to discovering optimal packet-dropping policies with reinforcement learning. Despite these remarkable advancements, no previous work has compiled these methods in the form of a survey article. This paper presents the first thorough documentation and analysis of ML-based algorithms for AQM, in which the strengths and limitations of each proposed method are evaluated and compared. In addition, a novel taxonomy of ML approaches based on methodology is also established. The review is concluded by discussing unexplored research gaps and potential new directions for more robust ML-AQM methods.

cs.NI↗

arXiv2024

Power Evaluation of IOT Application Layer Protocols

The Internet of Things has affected all aspects of daily life, and the number of IoT devices is increasing day by day. According to forecasts, the number of Internet of Things devices will reach one trillion devices by 2035. The increase in the number of devices connected to the Internet will cause various concerns. One of the most important concerns is the energy and power consumption of these devices. Although Internet of Things modules are low in energy consumption, their widespread and large-scale use has made the issue of power consumption become the most important challenge in this field. For this reason, it is necessary to use communication protocols that, in addition to establishing efficient communication, impose minimal power consumption on the network. In this paper, application layer protocols such as MQTT, MQTT-SN, CoAP, and HTTP are simulated using the tools available in the Contiki operating system, including COOJA and Powertrace, and they { are evaluated} and compared with each other in terms of power consumption. According to the simulations performed by the mentioned tools, the MQTT-SN protocol was the least consuming protocol in terms of power consumption. After that, the CoAP protocol is placed, and with a slight difference, the MQTT protocol, which consumes more than MQTT-SN. Finally, the HTTP protocol consumes the most power, which makes it unsuitable for communication in the Internet of Things

cs.NI↗

arXiv2023

DiCuPIT: Distributed Cuckoo Filter-based Pending Interest Table

Named data networking is one of the recommended {\color{red}architectures} for the future of the Internet. In this communication architecture, the content name is used instead of the IP address. To achieve this purpose, a new data structure is added to the nodes of named data networking which is called Pending Interest Table (PIT). Scalability, memory consumption, and integration are the significant challenges in PIT design {\color{red} as} it needs to be updated for each packet, and it saves the name of the packet. This paper introduces a new data structure for PIT called DiCuPIT. DiCuPIT is a distributed data structure for the PIT table, {\color{red} that works} based on the Cuckoo filter and can cover the three features as above-mentioned. {\color{red} By} implementing this PIT, {\color{red} the lookup} time shows {\color{red} a 36\% reduction} compared to the methods based on the Bloom filter and 40\% based on hash tables. Moreover, the memory consumption is reduced by 68\% compared to the hash tables-based mechanisms and 31\% compared to the methods based on the Bloom filter.

cs.NI↗

arXiv2023

Many-fields Packet Classification Using R-Tree and Field Concatenation Technique

Software-defined Networking is an approach that decouples the software-based control plane from the hardware-based data plane proposed for enterprise networks; OpenFlow is the most famous flexible protocol that can manage network traffic between the control and the data plane. Software-Defined Networking (SDN) requires up to 18 fields of the packets header to be checked against a big many-fields ruleset to categorize packets into flows, the process of categorizing packets into flows is called packet classification. Network switches process all packets belonging to the same flow in a similar manner by applying the same actions defined in the corresponding rule. Packet classification facilitates supporting new services such as filtering, blocking unsafe sites traffic, routing packets based on the packet's header information, and giving priority to specific flows. High-performance algorithms for many-field packet classification had been gained much interest in the research communities. This paper presents a new method to implement the many-fields packet classification of SDN flow table using Rectangle Tree (R-Tree). In this method, source and destination IP addresses from each flow table entry have been converted to a two-dimensional point. The remainders of the rule's fields have been concatenated into a single field by taking the most important bits with rules' ID in order to be inserted into the R-tree, for each rule an effective small binary flag is used to indicate the field's size, type, and ranges. Subsequently, searching is performed on the rectangle tree to find the matched rules according to the highest priority. In the simulation using the class-bench databases, the results show that this method achieves very good performance, classification speed and reduces the number of memory accesses significantly.

cs.NI↗

arXiv2023

Decomposition Theory Meets Reliability Analysis: Processing of Computation-Intensive Dependent Tasks over Vehicular Clouds with Dynamic Resources

Vehicular cloud (VC) is a promising technology for processing computation-intensive applications (CI-Apps) on smart vehicles. Implementing VCs over the network edge faces two key challenges: (C1) On-board computing resources of a single vehicle are often insufficient to process a CI-App; (C2) The dynamics of available resources, caused by vehicles' mobility, hinder reliable CI-App processing. This work is among the first to jointly address (C1) and (C2), while considering two common CI-App graph representations, directed acyclic graph (DAG) and undirected graph (UG). To address (C1), we consider partitioning a CI-App with $m$ dependent (sub-)tasks into $k\le m$ groups, which are dispersed across vehicles. To address (C2), we introduce a generalized reliability metric called conditional mean time to failure (C-MTTF). Subsequently, we increase the C-MTTF of dependent sub-tasks processing via introducing a general framework of redundancy-based processing of dependent sub-tasks over semi-dynamic VCs (RP-VC). We demonstrate that RP-VC can be modeled as a non-trivial semi-Markov process (SMP). To analyze this SMP model and its reliability, we develop a novel mathematical framework, called event stochastic algebra ($\langle e\rangle$-algebra). Based on $\langle e\rangle$-algebra, we propose decomposition theorem (DT) to transform the presented SMP to a decomposed SMP (D-SMP). We subsequently calculate the C-MTTF of our methodology. We demonstrate that $\langle e\rangle$-algebra and DT are general mathematical tools that can be used to analyze other cloud-based networks. Simulation results reveal the exactness of our analytical results and the efficiency of our methodology in terms of acceptance and success rates of CI-App processing.

cs.DC↗

arXiv2022

Performance Evaluation of Apache Spark MLlib Algorithms on an Intrusion Detection Dataset

The increase in the use of the Internet and web services and the advent of the fifth generation of cellular network technology (5G) along with ever-growing Internet of Things (IoT) data traffic will grow global internet usage. To ensure the security of future networks, machine learning-based intrusion detection and prevention systems (IDPS) must be implemented to detect new attacks, and big data parallel processing tools can be used to handle a huge collection of training data in these systems. In this paper Apache Spark, a general-purpose and fast cluster computing platform is used for processing and training a large volume of network traffic feature data. In this work, the most important features of the CSE-CIC-IDS2018 dataset are used for constructing machine learning models and then the most popular machine learning approaches, namely Logistic Regression, Support Vector Machine (SVM), three different Decision Tree Classifiers, and Naive Bayes algorithm are used to train the model using up to eight number of worker nodes. Our Spark cluster contains seven machines acting as worker nodes and one machine is configured as both a master and a worker. We use the CSE-CIC-IDS2018 dataset to evaluate the overall performance of these algorithms on Botnet attacks and distributed hyperparameter tuning is used to find the best single decision tree parameters. We have achieved up to 100% accuracy using selected features by the learning method in our experiments

cs.NI↗