To address the limitations of half-space coverage, low information transmission efficiency of RIS assisted batteryless IoT, a STAR-RIS assisted transmission scheme was proposed. Specifically, a STAR-RIS was deployed between the batteryless device (BD) and the access point, facilitating the uplink information transmission from the BD to the access point, thereby achieving full-space network coverage of the BD. To further enhance energy transmission efficiency, energy beamforming was designed at the power station. To this end, the system sum-rate maximization problems for both space division multiple access (SDMA) and non-orthogonal multiple access (NOMA) scenarios were investigated, with jointly optimization of the relection coefficients of the BD, the energy beamforming vector at the base station and the coefficient matrices of the STAR-RIS. For the SDMA scenario, an iterative algorithm based on the block coordinate descent (BCD) framework was proposed to solve the optimization problem. Similarly, to address the non-convexity of the optimization problem in the NOMA scenario, the objective function was transformed to simplify its solution, and another iterative algorithm based on the BCD framework was proposed. Numerical results show the superiorities of the proposed schemes over the benchmark schemes. Moreover, the utilization of the NOMA achieves better performance than the SDMA.

Aiming at the limitations in effectiveness and stealth of existing offline reinforcement learning(RL) data poisoning attacks, a critical time-step dynamic poisoning attack was proposed, perturbing important samples to achieve efficient and covert attacks. Temporal difference errors, identified through theoretical analysis as crucial for model learning, were used to guide poisoning target selection. A bi-objective optimization approach was introduced to minimize perturbation magnitude while maximizing the negative impact on performance. Experimental results show that with only a 1% poisoning rate, the method reduces agent performance by 84%, revealing the sensitivity and vulnerability of offline RL models in unmanned systems.

An augment negative example discrimination paradigm based on contrastive learning was proposed to improve the model’s fine-grained discrimination ability of video actions. The most challenging video-text negative pairs was generated, forming an augmented negative example set for each video sample. Based on this paradigm, a multi-granularity discrimination model for video action recognition was proposed to further distinguish between positive and negative examples. In this model, video features were extracted by the video representation module guided by textual positive examples, while self-correlation relationships between positive and negative semantics were established by the semantic discriminator equipped with a self-attention mechanism. Meanwhile, a coarse-grained distinction between the video modality and the augmented negative example set was achieved, while a fine-grained distinction between positive examples and the augmented negative example set within the text modality was also accomplished. Experimental results demonstrate that the augment negative set improves the model’s recognition ability on fine-grained class labels, and the multi-granularity discrimination model outperforms current state-of-the-art methods on the Kinetics-400, HMDB51 and UCF101 datasets.

Most methods do not model the pedestrian groups in autonomous driving, which will have an impact on road traffic safety. Therefore, a group perception pedestrian trajectory prediction network called GPCNet was proposed. Specifically, in intra-group, the interaction between pedestrian was learned at the individual level and the preference issue of different pedestrian was considered. In inter-group, the interaction between pedestrian groups was learned at the group level and the collision issue of pedestrian trajectory was considered using the social force model. Simulation results demonstrate that GPCNet improves the performance on the ETH and UCY datasets by 75.4% compared to the commonly used trajectory prediction methods.

Decentralized blockchain can realize the storage of medical data. Due to high isolation of blockchain from external environments, medical data in inside blockchain of medical institutions can only be shared within institution itself. To address the requester identity authentication and shared data security in cross-chain medical data sharing, a cross- chain medical data sharing scheme based on elliptic curve signcryption was devised in this article. Relay chain was used to realize interactive sharing of medical data. Cross-chain medical data was hierarchically processed for better effect of cross-chain data exchange. Signcryption process was completed by smart contracts to obtain high execution efficiency. Deletion and modification functions of medical data were deployed in smart contracts to solve the frequent updation problem of medical data.

To solve the problem of phased-array radar individual identification in the complex electromagnetic environment with wide spectrum, heterogeneous waveforms and strong energy, an unintentional modulation feature extraction method based on phase and frequency fusion was proposed for the individual features carried by phased-array radar signals. Considering that the phased array radar signal was difficult to collect, the number of transceiver components was large, and the unintentional modulation features were complex, the phased array radar unintentional modulation signal model was constructed based on the method of wave-position orchestration and isophase surface. Based on the bispectral method, the signal bispectral map was obtained and perimeter integration was performed to extract the unintentionally modulated phase features of the signal. Based on the variational modal decomposition method, the original signal was decomposed to obtain the modal components, and the energy ratio difference of the set of modal components was further computed to extract the unintentionally modulated frequency features of the signal. Finally, the local holding projection method was used to integrate the phase and frequency features, and the K nearest-neighbor classification method was adopted based on the tree retrieval method to realize the individual identification. Representative numerical results are reported, indicating that the proposed method has higher recognition accuracy and efficiency.

Existing anonymous communication systems that resisted traffic analysis could hide the identities of the communicating parties from the attacker. However, the identities of the communicating parties couldn’t be hidden from each other, and thus these systems did not apply to the scenario of anonymous whistleblowing and replying, where it was necessary to protect the identity of the whistleblower. To address this issue, an efficient anonymous whistleblower response scheme was proposed. With the technology of distributed point functions and secret sharing, the message was stored in two separate mailbox databases of non-colluding servers, so that the identity of the data receiver was hidden from the attacker. With the technology of secret sharing and encryption, the email address of the whistleblower was hidden, so that the receiving organization could reply without learning the whistleblower’s identity. The security analysis showed that the proposed scheme enabled the anonymity of both data receivers and whistleblowers at the same time. The experimental results show that compared to the Express scheme, the proposed scheme reduces the computational complexity during a reply to O(1) from O(logN), resulting in a 60% reduction in computational overhead for the receiving organization and a 50% reduction for the server.

To address the challenges of time-varying node connectivity and frequent link failures in tactical communication networks under conditions of strong adversarial and high-mobility operations, which result in frequent route convergence and short effective transmission times, a cluster-based route convergence method based on network state graph model (OSPF-CSG) was proposed. Firstly, link states were obtained based on the statistical characteristics of Hello messages, and node states were characterized using aggregated link state features to construct the network state graph model. Secondly, the neighbor state machine’s state transition triggers were improved, and two new OSPF message types were developed to design an active/passive association mechanism for complete adjacency relationships. Finally, on this basis, a cluster head bootstrap algorithm, node clustering algorithm, and inter-cluster connection algorithm were proposed to achieve cluster-based routing convergence. Simulation results show that, in eight different network topologies and under different link damage conditions, OSPF-CSG achieves an average reduction of 82% in the total number of route convergences, a reduction in routing overhead by 70% on average, and an improvement in packet delivery rate by 60% on average compared to the traditional OSPF protocol’s route convergence algorithm.

In the edge computing environment, there are risks such as communication channel risks and edge server failures, which can lead to a mismatch between the computing resources required for task processing and the resources allocated by the edge coalition. In response, a revenue forecasting method for the edge coalition and its member edge infrastructure provider (EIP) based on the game theory of uncertain coalition structures was proposed. Firstly, a resource scheduling model was constructed using a mixed integer linear programming method to maximize the revenue of the edge coalition. Secondly, a belief structure was introduced to characterize the probabilities of high, medium, low, and unknown scenarios for the coalition's revenue. Finally, the uncertain Owen value was used to estimate the interval revenue of the EIP in the coalition one time slot in advance. The simulation results show that the accuracy of this forecasting method under the two risks of channel risk and server failure is 91.25% and 82.5% respectively, with an average accuracy of 86.88%, achieving a relatively accurate forecast of the EIP’ revenue.

Issues of limited scene adaptability, inadequate evidence preservation, and low efficiency in traditional digital forensics were addressed by analyzing the feasibility of incorporating decentralized, tamper-resistant blockchain technology into digital forensic practices. Initially, a phased forensic process was proposed based on a hierarchical architecture for blockchain forensic technology, examining the advancements of blockchain at each stage of evidence acquisition, preservation, and presentation. Subsequently, limitations in existing research were analyzed, and a digital forensic framework incorporating comprehensive blockchain involvement was designed by utilizing the distributed advantages of blockchain. This framework integrated evidence information into the on-chain data structure and introduced a complementary graph analysis algorithm to standardize evidence collection across various scenarios. An off-chain distributed database was employed to achieve scalable, efficient storage, while smart contract templates enhance the reusability of contracts for similar forensic transactions. Lastly, potential future directions for the application of blockchain technology in forensic science were explored.

Aiming at the problem of optimal detection of mixed signal symbol sequences, on the basis of co-frequency mixed receiving baseband signal model, sufficient statistics for joint maximum likelihood blind detection were derived, and the optimal receiver design of “mixed waveform matched filtering and symbol rate dual sampling” was given. Based on the sufficient statistics model, a lower bound on the performance of single-channel blind demodulation for co-frequency mixed signals was derived using Forney’s method. Since the complexity of the optimal receiver grows exponentially with the equivalent symbol rate sampling channel length, the impact of the truncation error on the blind demodulation performance was further analyzed for the channel truncation problem in practical applications. The result shows that the simulation results of the optimal receiver coincide with the theoretical performance analysis results, verifying the reasonableness of the derived performance bounds.

In order to expand the existence space of enhanced cross Z-complementary set (E-CZCS), based on the extended generalized Boolean function (EGBF), the direct construction method of enhanced cross Z-complementary sets was proposed. The constructed q-ary E-CZCS had flexible and variable size, and their length was no longer limited to the traditional power of two. At the same time, the obtained enhanced cross Z-complementary sets was used as the training sequence of the generalized spatial modulation (GSM) system for channel estimation. The simulation results show that the enhanced cross Z-complementary sets can achieve optimal channel estimation. This presumably provides greater flexibility in the selection of training sequences in generalized spatial modulation systems.

To address the issues of frequent handovers and network load imbalance caused by dynamic changes in the network environment and enhanced mobility of user terminals in the 6G space-air-ground integrated network (SAGIN), a handover algorithm for SAGIN based on a terminal location prediction model was proposed. The algorithm constructed a long short-term memory (LSTM) network terminal location prediction model optimized based on the sparrow search strategy, improving the accuracy of terminal location prediction and resolving the issue of unreasonable handover timing. Based on this model, the SAGIN selection problem was modeled as a Markov decision process. A network handover algorithm utility function characterized by quality of service (QoS) requirements, handover cost, and network load balancing was designed. A distributional deep Q-network (D-DQN) was employed to select the network nodes that could maximize long-term goals for execution handover. Compared with network handover algorithms based on Q-Learning, double deep Q-network (DDQN), and dueling double deep Q-network (D3QN), the proposed algorithm performs better in terms of reducing handover delay and frequency, as well as enhancing network throughput, thereby validating the effectiveness of the proposed algorithm.

Program similarity analysis had a wide range of applications in areas such as code plagiarism and property protection, but it generally suffered from problems such as excessive computational overhead, a code similarity analysis method based on fuzzy matching and statistical inference was proposed. For binary programs, first disassembly analysis was performed and then function boundary recognition operations was performed to extract the execution boundary information of the function. On this basis, dynamic programming analysis methods were used to obtain similarity results between basic blocks at the granularity of the basic blocks, and neighborhood search was performed on the basis of the control flow graph to extend similarity analysis from the basic block level to the function level. Finally, the semantic similarity of binary files was obtained through statistical analysis of similarity functions. During this process, the pre trained model was optimized and analyzed, and the parameters were tuned to enable similarity analysis of cross platform code. The experimental results show that the proposed method has a significant improvement in analysis accuracy compared to traditional analysis tools, with an average increase of 7.1% in analysis accuracy compared to current mainstream analysis tools.

Federated learning as an emerging distributed machine learning, can solve the problem of data islands. However, due to the large-scale, distributed nature and strong autonomy of local clients, federated learning is extremely vulnerable to Byzantine attacks and the attacks are not easy to detect, which seriously damages the integrity and availability of the model. First, taking Byzantine attacks as the research object, a detailed classification and analysis of the attack principles were conducted. Secondly, guided by the classic network security defense model, federated learning defense methods were classified and analyzed from the perspective of defense mechanisms. Finally, the key issues and research challenges that need to be solved in federated learning to resist Byzantine attacks were proposed, providing new references for future relevant researchers.

The application and funding of the general programs, young scientists funds, funds for developing regions of National Natural Science Foundation of China (NSFC) in 2024 in information and communication system were statistically analyzed. Detailed statistics were provided among different institutions, professional titles, research topics and keyword frequencies. The key programs in terms of application and funding statistics as well as of different research topics were analyses. Moreover, the ages and professional titles of applicants for excellent and outstanding young scholar funds were also revealed, while the impact of the relevant NSFC reforms in 2024 on scholar funds was also analyzed. Finally, major programs and young student funds were briefly introduced, in order to provid landscape of NSFC in information and communication systems in order to let scientists prepare for applications in 2025.

Focusing on the issues of low efficiency and effectiveness in decision-making as well as the instability of traditional reinforcement learning model-based multi-function radar (MFR) jamming decision algorithms, a prior knowledge embedded long short-term memory (LSTM) network-proximal policy optimization (PPO) model based intelligent interference decision algorithm was developed. Firstly, the MFR interference decision problem was regarded as a Markov decision process (MDP). Furthermore, by incorporating prior knowledge associated with the interference domain into the reward function of the PPO model using revenue shaping theory, a reshaped reward function was obtained to guide agent converge quickly so as to improve decision-making efficiency. Besides, leveraging LSTM’s excellent temporal feature extraction ability enables capturing dynamic characteristics of echo data effectively to describe radar working states. Finally, these extracted dynamic features were inputted into the PPO model. With guidance from embedded prior knowledge, an effective interference decision can be achieved rapidly. Simulation results demonstrate that compared to traditional reinforcement learning model based interference decision algorithms, higher efficiency and effectiveness in decision-making can be attained via the proposed algorithms and the MFR interference decision can be efficiently and robustly achieved.

Vehicles positioned at varying locations in the platoon exhibit differing levels of energy consumption, which may result in the platoon breaking up during travel or even prevent successful formation of the platoon. To solve this problem, a vehicle platooning scheme based on two-way auction was proposed. Specifically, the positions in the vehicle platooning were considered as a special commodity to be auctioned. By requesting a price from the platoon leader and a bid from the platoon follower, the position of each vehicle participating in the platooning was determined. Furthermore, an incentive compatible platooning compensation mechanism was devised that could motivate the vehicles in the platoon to maintain their position while traveling. Theoretical analysis and extensive experiments show that, the proposed scheme can not only realize the vehicle platooning efficiently, but also ensure the stability of the formed platoon.