Abstract Autoencoders have been widely used in video anomaly detection, and there are many variants. However, since these Autoencoders use frames as the input of the reconstruction network, they can only learn the pixel information in the frame and lack other key information. In addition, complete reconstruction of the frame will lead to memory consumption and waste of resources. In order to learn multi‐dimensional information and reduce memory usage, we propose to use the Object‐meta instead of video frames, and the Memory Search Guided Autoencoder with Memory Pools (MSGAE‐MP) to reconstruct. Every Object‐meta comes from the object and is composed of the type mask, position mask, optical flow, and pixels of the object. In this way, the multi‐dimensional information carried by the input can be strengthened. After generating Object‐meta, the MSGAE‐MP will use the high‐dimensional information to guide search in memory modules during the reconstruction, and it will construct multi‐level memory pools, so as to reconstruct Object‐meta in different dimensions. Experiments show that our method is feasible and has achieved excellent results.

Link failures have a significant impact on network reliability. Reactive recovery schemes rely on rerouting but suffer from high latency. Alternatively, proactive recovery schemes require preselecting backup paths in advance, but they also face challenges such as expensive TCAM consumption on the data plane and delays in associating backup paths with network status changes. On that account, a link failure recovery method called LSNC (Link Failure Recovery through Long and Short Nested Cycles) is devised. At the beginning of a long cycle, LSNC classifies links and employs the DBPS (Dynamic Backup Path Selection) algorithm to select backup paths for each active link; and during each nested short cycles, LSNC updates the backup paths in the light of network status changes predicted by GRU. LSNC is a hybrid approach, combining elements of both reactive and proactive methods, thereby inheriting the advantages of each while refraining disadvantages. Experimental results show that the proposed LSNC method outperforms the known methods with respect to network load balancing and TCAM consumption.

Detecting Distributed Denial of Service (DDoS) attacks effectively has long been a challenge due to their disruptive nature. As the amount of network traffic increases, artificial intelligence based methods for detecting DDoS attacks have faced challenges of lower efficiency due to model training. On that account, a DDoS alerting method based on optimal sub-OD (Origin-Destination) pairs and Markov model (SODMM) is devised. SODMM first designs a subset feature evaluation technique to identify the most informative OD pairs, subsequently forming a TM based on these pairs. Then, the Generalized Network Temperature (GNT) derived from the TM is used to define the state transition matrix of a Markov model, dynamically adjusting such a transition matrix to enhance the accuracy of DDoS attack predictions. Experimental results show that SODMM outperforms the GNT, NAEW-GNT, and Rényi-GNT methods in terms of early alerts, possessing the highest ACC of 95.6%, as well as the lowest FPR of 10.06% and FNR of 3.83%.

To create network slices on a substrate network with limited resources, generally, an InP (internet provider) prioritizes the network revenue over the service fairness. However, such a strategy may negatively impacts the InP’s market share in the long run, because a key to a successful and sustainable business is to better serve all customers with limited resources. It is noteworthy that increasing both fairness and revenue is challenging as these two requirements are at odds. To effectively increase fairness while minimizing decreases in revenues because of fairness, we propose a deep reinforcement learning (DRL) based approach, referred to as WAC-NIE (Weighted Admission Control & Node Importance-based network slice Embedding), which jointly optimizes network slice admission control and network slice embedding.

Online Social Networks (OSNs) face escalating security threats that imperil user privacy. Conventional Deep Learning methods, relying predominantly on fixed learning rates, encounter limitations when capturing the nuanced intricacies of OSN traffic that arise from shifting user behaviors, diverse content types, and evolving interaction patterns because of social trending topics changes. To tackle these challenges, our paper delves into the diverse variations and transitions from a uniform approach, where a single method is employed for various types of data, to a multi-variation methodology. This methodology dynamically adapts to the special characteristics of each data type, resulting in more effective data representation while alleviating the limitations associated with fixed-rate calibration. Therefore, we devise the Adaptive Swarm Reinforcement Learning (ASRL) method that leverages adaptive learning to intricately analyze a wide range of user interactions, endowing our proposed method with the capacity to flexibly adjust to the constantly shifting OSN patterns. The experiments show that the proposed ASRL method achieves an accuracy of 98.59% in detecting a range of threat patterns, surpassing other prevalent methods by an average of 5% across the datasets from Facebook, Google+, and Twitter. Meanwhile, ASRL logs suspicious activities to identify the intruder for forensic analysis. The implementation of our proposed method is now publicly accessible athttps://github.com/don2c/asrl_Project.

A social network has an in-born core-fringe structure. To increase the core nodes resolution, the paper proposes a new method, named KSCNR (K-Shell and Salton index based core node recognition) method, that combines both the local network topology features (Salton index with gravitational centrality) and the global network topology features (K-Shell iteration) to identify core nodes. The KSCNR method utilizes the weights to adjust the influences of the local and the global topology features according to the core nodes preferences, which makes the KSCNR method suitable for different social network scenarios. The experimental results show that the KSCNR method outperforms the known methods such as the K-Shell, the BC, the DC and the CC methods in the light of both effectiveness and accuracy.

The major threat to online social network (OSN) users is account compromisation. Spammers now spread malicious messages by exploiting the trust relationship established between account owners and their friends. The challenge in detecting a compromised account by service providers is validating the trusted relationship established between the account owners, their friends, and the spammers. Another challenge is the increase in required human interaction with feature selection. Research available on supervised learning has limitations with feature selection and accounts that cannot be profiled, like application programming interface (API). Therefore, this article discusses the various behaviors of OSN users and the current approaches in detecting a compromised OSN account, emphasizing its limitations and challenges. We propose a deep learning approach that addresses and resolve the constraints faced by previous schemes. We detailed our proposed optimized nonsymmetric deep autoencoder (OPT_NSDAE) for unsupervised feature learning, which reduces the required human interaction levels in the selection and extraction of features. We evaluated our proposed classifier using three different social network datasets, Facebook, Google+, and Twitter, in addition to the NSL-KDD and KDDCUP’99 datasets, in a graphical-user-interface-enabled Weka application. The experimental results show that our proposed approach outperformed most of the traditional schemes in OSN compromised account detection.

In wireless sensor networks (WSNs), provenance is critical for assessing the trustworthiness of the data acquired and forwarded by sensor nodes. Due to the energy and bandwidth limitations of WSNs, it is crucial that data provenance should be as compact as possible. The main drawback of the existing block provenance schemes is that to decode the provenance, all of the provenance blocks must be received by the base station (BS) correctly. To address such an issue, we propose a multigranularity graphs-based stepwise refinement provenance scheme (MSRP), among which we use the mutual information between node pair as the similarity index to classify node IDs and then generate the multigranularity topology graphs. Furthermore, the dictionary-based provenance scheme (DP) is employed to encode the provenance in a stepwise manner. The BS recovers the provenance in the same stepwise manner and performs the data trustworthiness evaluation during the decoding. We evaluate the performance of the MSRP scheme extensively by both simulations and testbed experiments. In addition to mitigating the main drawback of the existing block provenance schemes, the results show that our scheme not only outperforms the known related schemes with respect to average provenance size and energy consumption but also drastically improves data trustworthiness assessing efficiency.

In a wireless sensor network (WSN), provenance, usually considered as the tracebacks of the data packets’ acquisition and transmission, is critical for assessing data trustworthiness. However, the provenance size expands rapidly with increases in the number of packet transmission hops. Among the known provenance schemes, the dictionary based provenance (DP) scheme achieves the highest provenance compression rate up to now. Nevertheless, the average compression rate of the DP scheme decreases drastically when the WSN’s topology is not stable. To address such a disadvantage, in this paper we propose a packet path index differences based provenance (PIDP) scheme. In the PIDP scheme, the backbone paths along the gradient directions of the packet transmissions are built, and then the provenance is encoded by a selected index of a backbone path dictionaries together with a deviation from the backbone path. To further increase the provenance compression rate and decrease the computation complexity at each node, a packet path hash value based provenance (PHP) scheme is proposed as a complementary approach of the PIDP scheme, in which the provenance is encoded as the data source node ID together with a segment of the packet path’s hash value. Both the simulation and experimental results show that our proposed two schemes, the PIDP and the PHP schemes, outperform the DP scheme with respect to both provenance compression rate and energy conservation rate even when a WSN’s topology is not stable.

Being able to monitor each packet path is critical for effective measurement and management of networks. However, such detailed monitoring can be very expensive especially for large-scale networks. To address such problem, inspired by thermodynamics, which uses the statistical characteristics of a large number of molecules’ motion but not each molecule’s trajectory for analysis, we propose the new concept of network temperature together with the notions of network-specific heat and network temperature gradient . Our approach does not only provide a statistical view of the current network state consisting of all the active packet paths at each time instant, but can be used to represent transitions among network states. Our network temperature-based methods have a broad applicability, such as to DDoS detection, dynamic node importance ranking, network stability and robustness evaluation, reliable packets routing, provenance compression assessment, and so on. Numerical and/or the experimental results show that our methods are effective.

Traffic anomaly detection is a key research topic for large scale communication networks. Traditional network entropy has been proved to be an effective metric on network traffic anomaly detection. However, such a method also shows limitations in large scale networks with constantly changing packet flows, which makes the traditional entropy based method inefficient for traffic anomaly detection. To address this problem, we propose a novel indicator named Entropy Change Rate to improve the effectiveness of the traditional entropy based network traffic anomaly detection.

A new indoor location method based on multi-parameter access point selection and device heterogeneity weakening is proposed. The Dixon Criterion is applied to filter and select AP through the three parameters such as the Wi-Fi signals stability, the signals relevance, and the reference point discrimination ability. The Wasserstein distance is applied as a weight in our positioning algorithm to mitigate the positioning errors caused by device heterogeneity. The experimental results show that with the proposed method the average positioning error is reduced by 10% and the positioning stability is improved by 94.9% compared with the known WKNN method.

When a network is suffering from attack and cannot afford to protect all the nodes, it is reasonable to pick up the important nodes and then protect them with priorities. Many node importance ranking methods have been proposed, but most of them are only based on static network topology analysis and rarely take the impact of dynamic network traffic into account. As a result, we devise a novel method named MixR (Mix Ranking) algorithm which takes both network topology and dynamic network load into account to rank the nodes importance. To show the effectiveness of MixR algorithm, we apply SIR (Susceptible-Infected-Recovered) as an inference model to compare our proposed method with the other known node importance ranking methods such as Degree Centrality(DC), Closeness Centrality(CC), Eigenvalue Centrality(EC) and Semi-local Centrality. The experimental results show that MixR algorithm outperforms those known methods with respect to dynamic network traffic changes in the network.

Information Communication Technology (ICT) environment in traditional power grids makes detection and mitigation of DDoS attacks more challenging. Existing security technologies, besides their efficiency, are not adequate to cater to DDoS security in Smart Grids (SGs) due to highly distributed and dynamic network environments. Recently, emerging Software Defined Networking- (SDN-) based approaches are proposed by researchers for SG’s DDoS protection; however, they are only able to protect against flooding attacks and are dependent on static thresholds. The proposed approach, i.e., Software Defined Networking-based DDoS Protection System (S-DPS), is efficiently addressing these issues by employing light-weight Tsallis entropy-based defense mechanisms using SDN environment. It provides early detection mechanism with mitigation of anomaly in real time. The approach offers the best deployment location of defense mechanism due to the centralized control of network. Moreover, the employment of a dynamic threshold mechanism is making detection process adaptive to the changing network conditions. S-DPS has demonstrated its effectiveness and efficiency in terms of Detection Rate (DR) and minimal CPU/RAM utilization, considering DDoS protection focusing smurf attacks, socket stress attacks, and SYN flood attacks.

Wireless sensor networks (WSNs) are employed for different applications for the reason of small‐sized and low‐cost sensor nodes. However, several challenges that include a low powered battery of the sensor nodes restrict their functionality. Therefore, saving energy in the routing process to extend network life is a serious concern while deploying applications on WSN. To this end, the key technology is clustering, which helps maximize scalability and network lifecycle. Base station (BS) collects data, aggregates it, and extracts the required information. To obtain the maximum outcome, the lifetime of the network is maximized by the use of different techniques and protocols. Data transmissions consume most of the network energy, and the transmissions over normal ranges require less energy as compared to transmissions over long ranges. Moreover, the nodes closer to the BS deplete their energy faster as compared to distant nodes because of traffic overload. The proposed protocol is aimed at reducing energy consumption and increasing the network lifetime. For this purpose, the network is divided into two regions: region 1 closer to the BS communicating directly, whereas region 2 farther away from the BS having routing nodes to communicate with the BS. Routing nodes do not take part in sensing function but will only move in region 2 collecting data and forwarding it to BS. MATLAB is used as the simulation tool for evaluation, and the results are compared with the existing optimized region‐based efficient routing (AORED) and low‐energy adaptive clustering hierarchical protocol (LEACH) techniques. The comparison showed that energy conservation and lifetime increased by 15%, and throughput is increased by more than 5% approximately.

In wireless sensor networks (WSNs), provenance is critical for assessing the trustworthiness of data acquired and forwarded by sensor nodes, detecting early signs of attacks, etc. However, the provenance size expands rapidly with increases in the number of packet transmission hops. Among the existing provenance schemes, the dictionary based provenance scheme (DP) achieves the highest provenance compression rate. However, the major drawback of the DP scheme is that it is sensitive to the WSN's topology changes, which cannot be used in the WSNs with rapid topology changes. To overcome such a drawback and achieve a higher compression rate, we propose a path index differences based provenance scheme, in which we first establish backbone paths along the gradient direction, and then we devise a Truncation Hamming Distance (THD) based method to eliminate the backbone paths with high similarity and build the path dictionaries for the selected backbone paths of low similarity. With the support of such dictionaries, a new path is encoded by the index of a similar path in the dictionary together with the differences between them, which makes the size of the provenance stably compressed. Compared to the DP scheme, the simulation and experimental results show that our scheme can achieve a higher provenance compression ratio even if the topology structure of the WSN is not stable.

Since the classical hash function widely used are costly for resource constrained devices, in this paper we propose a new lightweight hash function LNHASH through Cellular Automata(CA), in which sponge construction is applied as the mode of operation. Both linear and non-linear CA rules are employed to construct the internal permutation to achieve high diffusion and confusion. LNHASH allows making trade-offs among security, speed, energy consumption and implementation costs by adjusting the corresponding parameters. The hardware-friendly Sbox and linear layer bring low area implementation. Security analysis results show the resistance of the LNHASH scheme against the known attacks.

In wireless sensor networks (WSNs), provenance is vital for assessing data's trustworthiness, detecting the misbehaviors conducted by adversaries or troubleshooting communication failures. The provenance can be encoded through fingerprinting the node IDs along a packet path where the packets are generated, forwarded and/or aggregated. Because WSNs are resource-tightened networks, most of the known provenance schemes applied in WSNs address the issues on how to reduce the provenance size with various compression techniques only. However, reducing the provenance size at a sensor node also costs too much energy. In addition, such schemes did not take the secure and persistant provenance storage for consideration in a long term. To fill the gap, we propose a blockchainbased data provenance scheme (BCP) of compression free, where the provenances are stored distributively on the nodes along the packet path and the BS can retrieve the provenance on demand through a query process. An edge computing based monitor network consisting of high performance nodes (H-nodes) is deployed above or nearby the WSNs, which keeps the WSN's provenance data in a blockchain-based database. The security and authenticity of the provenances are then protected. What's more, the WSN is released from consuming much energy in handling provenance data, which is more superior to all the previous schemes. Both the simulation and experiment results show that our scheme BCP is more energy efficient and secure than those of the known distributed data provenances.

In today’s world we are surrounded by world of smart handheld devices like smart phones, tablets, netbooks, and others. These devices are based on advance technologies of multiple‐input and multiple‐output, Orthogonal Frequency Division Multiplexing (OFDM), and advance data reliability techniques such as forward error corrections. High data rates are among the requirements of these technologies for which turbo and low density parity check codes (LDPC) are widely used in these standards. In order to get high speed, we need multiple and parallel processors for the implementation of such codes. But there exists a collision problem as a consequence of parallel processor. This problem results in increase latency and increase of hardware complexity. In this work an approach for collision problem is presented in which network relaxation technique is used which is based on a fast clique detection. The proposed approach results in high throughput in terms of latency and complexity. Furthermore, the proposed solution is able to solve the collision problem by connecting network optimization for achieving high throughput.

In wireless sensor networks (WSNs), data provenance records the data source and the forwarding and the aggregating information of a packet on its way to the base station (BS). To conserve the energy and wireless communication bandwidth, the provenances are compressed at each node along the packet path. To perform the provenances compression in resource‐tightened WSNs, we present a cluster‐based arithmetic coding method which not only has a higher compression rate but also can encode and decode the provenance in an incremental manner; i.e., the provenance can be zoomed in and out like Google Maps. Such a decoding method raises the efficiencies of the provenance decoding and the data trust assessment. Furthermore, the relationship between the clustering size and the provenance size is formally analyzed, and then the optimal clustering size is derived as a mathematical function of the WSN’s size. Both the simulation and the test‐bed experimental results show that our scheme outperforms the known arithmetic coding based provenance compression schemes with respect to the average provenance size, the energy consumption, and the communication bandwidth consumption.

Provenance records the history of data acquisition and transmission. In wireless sensor networks (WSNs), provenance is critical for many different purposes, including assessing the trustworthiness of data acquired and forwarded by sensors, supporting situation awareness, and detecting early signs of attacks. However, a major drawback in provenance for WSNs is its size. It is thus critical to develop efficient techniques for provenance encoding. A major issue of previously proposed provenance encoding techniques is that the size of the provenance either expands too fast with increases in the number of packet transmission hops or is very sensitive to the WSN’s topology, i.e., the size of the provenance expands drastically with changes in the WSN’s topology. In this article, we propose a novel provenance encoding technique based on dynamic Bayesian network and overlapped arithmetic coding scheme, which addresses such an issue. Through theoretical analysis, simulation, and testbed experiments, we show that our scheme outperforms other WSN lightweight provenance schemes with respect to provenance size and energy consumption.

In wireless sensor networks (WSNs), provenance records the data source, forwarding, and aggregating information of data packets on their way to the base station. Provenance is critical for assessing the trustworthiness of the received data, diagnosing network failures, detecting early signs of attacks, etc. However, because the provenance size expands rapidly with the increase in packet transmission hops, the provenance schemes developed for use in wired computer networks are not generally applicable to WSNs. Therefore, specific provenance techniques have been developed for WSNs that take into account the constrained resources of sensor nodes. In this paper, we survey such techniques. Special focus in the paper is devoted to a systematic and comprehensive classification of the solutions proposed in the literature. We review each solution by highlighting its pros and cons. Finally, we discuss recent trends in provenance encoding schemes for WSNs.

Due to energy and bandwidth limitations of wireless sensor networks (WSNs), it is crucial that data provenance for these networks be as compact as possible. Even if lossy compression techniques are used for encoding provenance information, the size of the provenance increases with the number of nodes traversed by the network packets. To address such issues, we propose a dictionary based provenance scheme. In our approach, each sensor node in the network stores a packet path dictionary. With the support of this dictionary, a path index instead of the path itself is enclosed with each packet. Since the packet path index is a code word of a dictionary, its size is independent of the number of nodes present in the packet's path. Furthermore, as our scheme binds the packet and its provenance through an AM-FM sketch and uses a secure packet sequence number generation technique, it can defend against most of the known provenance attacks. Through simulation and experimental results, we show that our scheme outperforms other compact provenance schemes with respect to provenance size, robustness, and energy consumption.

Since data are originated and processed by multiple agents in wireless sensor networks, data provenance plays an important role for assuring data trustworthiness. However, the size of the provenance tends to increase at a higher rate as it is transmitted from the source to the base station and is processed by many intermediate nodes. Due to bandwidth and energy limitations of wireless sensor networks, such increasing of provenance size slows down the network and depletes the energy of sensor nodes. Therefore, compression of data provenance is an essential requirement. Existing lossy compression schemes based on Bloom filters or probabilistic packet marking approaches have high error rates in provenance-recovery. In this paper, we address this problem and propose a distributed and lossless arithmetic coding based compression technique which achieves a compression ratio higher than that of existing techniques and also close to Shannon's entropy bound. Unlike other provenance schemes, the most interesting characteristic of our scheme is that the provenance size is not directly proportional to the number of hops, but to the occurrence probabilities of the nodes that are on a packet's path. We also ensure the confidentiality, integrity, and freshness of provenance to prevent malicious nodes from compromising the security of data provenance. Finally, the simulation and testbed results provide a strong evidence for the claims in the paper.