Internet of Things (IoT) applications continuously generate large volumes of transient data. Delivering transient data efficiently is challenging because it is short-lived, highly dynamic, and often critical for time-sensitive services. Caching at the Edge offers a practical solution by storing frequently requested content closer to users, reducing delivery delays, and easing network congestion. However, existing caching approaches in Edge-assisted IoT networks face four significant limitations: (i) lack of freshness-aware policies, leading to outdated data, (ii) static or centralized coordination, which restricts scalability, (iii) inability to adapt to bursty and heterogeneous traffic patterns, and (iv) inefficient handling of resource-constrained Edge nodes. IoT-Cooperative Caching (IoT-C) addresses these issues with a framework based on multi-agent reinforcement learning. Using the framework, Edge servers make decentralized, adaptive decisions that account for both user demand and data freshness. IoT-C introduces topic-based grouping of Edge nodes and a hierarchical state model that supports collaboration across local, group, and global states. Experiments show that IoT-C increases cache hit rates, reduces latency, and improves freshness compared with state-of-the-art techniques. These improvements make the proposed approach well-suited for time-critical IoT applications like smart cities, healthcare, and industrial networks.

In today’s technological world, the rapid proliferation of Android malware presents significant security challenges in the mobile ecosystem. Mobile security researchers face mounting challenges from increasingly sophisticated Android malware, highlighting critical limitations in conventional detection methods. While existing approaches struggle to process high-dimensional data and handle imbalanced sample distributions effectively, we address these shortcomings through a framework. By combining bio-inspired optimization techniques with modern graph-based neural architectures, this paper proposes DroidSwarmAI for classification. Using the Drebin and TUANDROMD datasets, our method effectively handles imbalanced data and high-dimensional features. Our experimental implementation incorporates specialized data preparation techniques, including targeted sampling strategies to address class distribution issues. We evaluate consistent datasets extensively, including adaptation analysis, computational cost comparison, and baseline reproduction. Our model achieves a peak accuracy of 98.63%, maintaining robustness against distribution drift and outperforming several state-of-the-art baselines.

The increasing prevalence of Android malware poses significant security risks, demanding efficient detection methods. Existing antivirus software uses signature-based detection, which often fails to identify zero-day attacks. Many existing Machine Learning (ML) based detection models are trained on millions of redundant features or use a single category of features to develop a detection model. Our work demonstrates the importance of systematic feature selection and combining features to improve the detection accuracy and reduce the computational overhead. We propose a three-level feature selection framework comprising Gradient Boosting (GB) to identify the top 100 features at the first level based on importance score to reduce the loss function, Recursive Feature Elimination (RFE) to refine these top 100 to the top 50 at the next level by removing features with the lowest importance score in each step until a predefined number of features remain, and a Voting Classifier (VC) ensemble for final classification by combining predictions from multiple base models. We train various ML models on feature categories such as Permissions, Activities, Services, API Calls, and Opcodes. Further, we take various combinations of these features to evaluate the effectiveness of hybrid feature vectors and our feature selection pipeline. The experimental results indicate that the proposed feature selection pipeline reduced the feature vector dimensions significantly and reduced training and detection latency. Furthermore, combining various categories of features improved the detection accuracy and reduced false alarms.

In the era of ubiquitous computing devices, malware is the primary weapon of cyber attacks, and malware-related security breaches remain a significant security concern. Nowadays, adversaries require fewer resources to exploit a system with the help of contemporary malicious payloads and AI tools than in the old days. Despite many advances in malware defense research, adversaries continually employ sophisticated tools and techniques to evade existing defense mechanisms and create chaos. Moreover, it is challenging to recognize these malicious binaries with shallow features such as section names, entropies, virtual sizes, and strings, which are not robust. The proposed work mainly focuses on identifying robust features that can help to detect more sophisticated (i) seen and (ii) never-seen-before malware effectively. Unlike the existing research works,$D^{2}4D$concentrates on four types of analysis: Registry key, API function, network, and memory analysis. Above all,$D^{2}4D$identifies the binaries that perform fast-flux attacks, DGA-based attacks, homoglyphs attacks, and other attack types. The evaluation results indicate that the$D^{2}4D$achieves an accuracy of 99.67%, with a 0.10% False Positive Rate for seen binaries and more than 91% accuracy for never-seen-before binaries. Beyond that,$D^{2}4D$outperforms 33 existing anti-malware. The extracted features prove robust in identifying seen and never-seen-before binaries based on the experimental analysis, comparison with the state-of-the-art models, and ablation study.

The monitoring of ambient environments relies heavily on data fetched from sensors in the Internet of Things (IoT) services. However, with the increasing number of mobile users and IoT applications, there has been a surge in traffic on IoT networks, leading to faster draining of sensor batteries. Hence, caching at the IoT Edge has emerged as a promising solution, reducing network congestion and energy consumption. This paper proposes an optimal custom caching strategy tailored to Edge computing. By considering factors such as the number of requests, battery level, and Age of Information (AoI) for each sensor, the Edge node decides whether to command a sensor to send a status update or retrieve data from the cache. We formulate the dynamic content caching challenge as a Markov Decision Process (MDP) to optimize jointly long-term caching costs. The proposed relative value iteration algorithm effectively solves the MDP problem without prior knowledge of user preference. Simulation demonstrates that the proposed policy outperforms the greedy and LRU policies with 10.61% and 44.05% lower cache miss rates, respectively, and a 69% slower energy depletion rate and significantly longer node longevity.

Publish-Subscribe (Pub-Sub) network is a communication paradigm where publishers produce data on specific topics and subscribers subscribe to these topics. Brokers are middleware facilitating data delivery and communication based on topics. However, brokers are geographically distributed in a multi-broker environment, offering different service times, and tend to become overloaded due to the popularity of the topics when an event occurs. This results in an intolerable average delivery delay. Therefore, we propose an efficient topic-aware low latency load balancing approach that reduces the delay by handling unequal traffic distribution due to data popularity, varied request rates, and hierarchical topic ordering. The proposed approach is manifold. First, it associates the topics to a broker using a Trie data structure, identifies Hot topics based on the outlier detection method, and balances the broker load based on sojourn time modeling and heuristic approach. We use a Pub-Sub MQTT testbed to experiment with a Mosquitto broker distributed network. It reduces the clients’ average waiting time by 11%. Our approach distributes the load 20% more evenly than other approaches, and the average server utilization rate is close to 22%, which is proximal to the optimal approach and better than other approaches under study.

Explosive traffic and service delay are bottlenecks in providing Quality of Service (QoS) to the Internet of Things (IoT) end-users. Edge caching emerged as a promising solution, but data transiency, limited caching capability, and network volatility trigger the dimensionality curse. Therefore, we propose a Deep Reinforcement Learning (DRL) approach, named IoT-Cache, to caching action optimization. An appropriate reward function is designed to increase the cache hit rate and optimize the overall data-cache allocation. A practical scenario with inconsistent requests and data item sizes is considered, and a Distributed Proximal Policy Optimization (DPPO) algorithm is proposed, enabling IoT edge nodes to learn caching policy. RLlib framework is used to scale the training in distributed Publish/Subscribe network. The performance evaluation demonstrates a significant improvement and faster convergence for IoT-Cache cost function, a trade-off between communication cost and data freshness over existing DRL and baseline caching solutions.

With the rapid increase in the adaptability of Internet of Things (IoT) based solutions worldwide, there is a dire need to implement these solutions effectively and efficiently. One step towards such implementation is to run Wireless Sensor Networks (WSN) or IoT devices on minimal energy with an uninterrupted power supply to maintain the system unattended. In a way, a self-sustainable solution is needed to meet the long-term energy requirements of the system. A rational design of IoT/WSN nodes is an essential criterion for the system to be energy efficient. In this paper, we create an IoT/WSN simulation environment in MATLAB and integrate it with clustering algorithms and energy harvesting schemes to improve the life expectancy of the nodes. The simulation environment includes an infrastructure with 100 IoT/WSN nodes to test the impact of several clustering algorithms like K-Means, K-Medoids, and Fuzzy C-Means on Wireless Energy Harvesting (WEH) schemes. The simulation results show that the clustering algorithms balance the energy dissipation from the nodes in a WSN/IoT network. The Energy Harvesting (EH) provides extra energy to cater to the IoT/WSN operation for an extended period. As a result, when the EH scheme is implemented with a clustering algorithm, there is a significant improvement in node life expectancy over the simulation period. The results show that we can effectively extend the network lifetime by merging these two approaches, leading to a long uninterrupted operation of the deployed infrastructure.

The Android smartphones are highly prone to spreading the malware due to intrinsic feebleness that permits an application to access the internal resources when the user grants the permissions knowingly or unknowingly. Hence, the researchers have focused on identifying the conspicuous permissions that lead to malware detection. Most of these permissions, common to malware and normal applications present themselves in different patterns and contribute to attacks. Therefore, it is essential to find the significant combinations of the permissions that can be dangerous. Hence, this paper aims to identify the pairs of permissions that can be dangerous. To the best of our knowledge, none of the existing works have used the permission pairs to detect malware. In this paper, we proposed an innovative detection model, named PermPair, that constructs and compares the graphs for malware and normal samples by extracting the permission pairs from the manifest file of an application. The evaluation results indicate that the proposed scheme is successful in detecting malicious samples with an accuracy of 95.44% when compared to other similar approaches and favorite mobile anti-malware apps. Further, we also proposed an efficient edge elimination algorithm that removed 7% of the unnecessary edges from the malware graph and 41% from the normal graph. This lead to minimum space utility and also 28% decrease in the detection time.

Permissions and the network traffic features are the widely used attributes in static and dynamic Android malware detection respectively. However, static permissions cannot detect stealthy malware with update attacks capability, while dynamic network traffic cannot detect the malware samples without network connectivity. Hence, there is a need to build a hybrid model combining both these attributes. In this work, we propose a hybrid malware detector that examines both the permissions and the traffic features to detect malicious Android samples. The proposed approach is based on the combination of Supervised Learning (KNN Algorithm) and Unsupervised Learning (K-Medoids Algorithm). Experimental results demonstrate that hybrid approach gives the overall detection accuracy of 91.98%, better than static and dynamic detection accuracies of 71.46% and 81.13% respectively.

Internet-of-Things (IoT) is emerging as one of the popular technologies influencing every aspect of human life. The IoT devices equipped with sensors are changing every domain of the world to become smarter. In particular, the majorly benifited service sectors are agriculture, industries, healthcare, control & automation, retail & logistics, and power & energy. The data generated in these areas is massive requiring bigger storage and stronger compute. On the other hand, the IoT devices are limited in processing and storage capabilities and can not store and process the sensed data locally. Hence, there is a dire need to integrate these devices with the external data sources for effective utilisation and assessment of the collected data. Several existing well-known message exchange protocols like Message Queuing Telemetry Transport (MQTT), Advanced Message Queuing Protocol (AMQP), and Constrained Application Protocol (CoAP) are applicable in IoT communications. However, a thorough study is required to understand their impact and suitability in IoT scenario for the exchange of information between external data sources and IoT devices. In this paper, we designed and implemented an application layer framework to test and understand the behavior of these protocols and conducted the experiments on a realistic test bed using wired, wifi and 2/3/4G networks. The results revealed that MQTT and AMQP perform well on wired and wireless connections whereas CoAP performs consistently well and is less network dependent. On the lossy networks, CoAP generates low traffic as compared to MQTT and AMQP. The low memory footprint of MQTT and CoAP has made them a better choice over AMQP.

The advent of Cloud has aided mobile devices in performing computation intensive tasks with the virtue of offloading. However, this leads to communication with the Cloud, which results in high energy consumption. Moreover, communication over wireless medium has the risk of intermittent connectivity with Cloud. Hence, there is an urgent need for providing a trade-off between energy consumption and service availability in Mobile Cloud Computing. This paper model the trade-off problem as a multi-criteria decision making optimization problem, considering different parameters such as energy consumption, waiting time, risk for offloading and deadline to compute a task. The proposed model takes intelligent offloading decisions to avail service from a remote resource such as Cloud. The validity of decision is tested with different applications and proved that the model conserves energy and also prolongs the service connection for mobile devices.

Close monitoring, proper control and management of plant diseases are essential in the efficient cultivation of crops. This paper presents a scheme that uses mobile phones for real-time on-field imaging of diseased plants followed by disease diagnosis via analysis of visual phenotypes. A threshold based offloading scheme is employed for judicious sharing of the computational load between the mobile device and a central server at the plant pathology laboratory, thereby offering a trade-off between the power consumption in the mobile device and the transmission cost. The part of the processing carried out in the mobile device includes leaf image segmentation and spotting of disease patch using improved k-means clustering. The algorithm is simple and hence suitable for Android based mobile devices. The segmented image is subsequently communicated to the central server. This ensures reduced transmission cost compared to that in transmitting full leaf image.

This paper presents a fast and efficient hand gesture based mobile augmented reality (MAR) system for interactive classroom. It provides a complex visual augmented layer over static slides to understand the concepts more clearly without touching the computer devices or using whiteboard. Simple hand gestures are used to interact with slides while presenting in the classroom or in any conference room with high accuracy and efficiency without any expensive hardware. The gesture path is tracked continuously using a color tracking algorithm proposed. A decision tree is used to make the decisions based on the gestures. The preliminary result indicates that the gesture recognition rate is near about approximately 94% and it is mostly acceptable. This enhances the user's interaction level with immersive feeling in immersive environment.