Yueheng Mao, Min Yu, Gengwang Li, Jianguo Jiang, Gang Li, Meng Zhang, Zhen Xu, Weiqing Huang, Ming Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Deepfake detectors have demonstrated vulnerability when faced with adversarial attacks. In real-world scenarios, attackers can generate adversarial examples that mislead detectors through query-based black-box attack methods. Recently, several defense methods against query-based adversarial attacks have achieved promising performance. However, their effectiveness significantly declines when applied to deepfake detection tasks. In this work, we propose a novel defense method specifically designed to counter adversarial attacks in deepfake detection scenarios. Unlike existing approaches, our method focuses on the high-frequency components of images to uncover subtle traces left by adversarial perturbations. Specifically, we analyze the high-frequency residual between similar images within the queries to detect adversarial examples. We evaluate our method against three advanced black-box attack strategies. Extensive experiments demonstrate that our approach achieves highly effective attack detection while significantly reducing FPR.

Spread spectrum communication plays a vital role in safeguarding the Internet of Things systems, due to its inherently low probability of interception and anti-jamming capability. However, conventional spread spectrum systems based on pseudorandom sequences are disadvantageous in limited sequence length and deterministic periodicity, making them vulnerable to brute-force attacks. To address these limitations and enhance the randomness of the spread spectrum sequences, a novel wireless channel randomness integrated spread spectrum sequence generation method is proposed in this work. Taking advantages of the intrinsic randomness, temporal variations, and unpredictability of wireless channel fadings, the proposed approach converts the extracted channel features into ordered sequences, which are then used to control the selection of irreducible generating polynomials for spread spectrum sequence generation. The proposed method improves the randomness and secrecy of the integrated spread spectrum sequence. Theoretical analysis and simulation results demonstrate that the proposed sequences not only achieve higher randomness entropy compared to the traditionalm-sequence, but also pass the National Institute of Standards and Technology randomness tests. Furthermore, performance evaluations under various signal-to-interference ratio conditions show improved autocorrelation properties and largely lower bit error rates, validating the effectiveness of the proposed method in improving the anti-jamming capability.

In recent years, the popularity of LEO satellite internet has made satellite security the focus of industry and academia, such as jamming and spoofing attacks aimed at physical satellite signals. Therefore, developing sophisticated anti-jamming and anti-spoofing technologies is essential, and individual identification of satellites is a prerequisite for implementing this technology. However, off-the-shelf LEO system signals exhibit significant differences in fading characteristics, communication protocols, modulation schemes, and Doppler shifts compared to terrestrial signals, which challenges current individual identification. Besides, individual identification research on off-the-shelf LEO systems such as Starlink remains relatively scarce. This paper proposes SatTransformer, a novel satellite signal individual identification method for LEO system signals, integrating nonlinear mapping and Vision Transformer techniques. We enhanced signal spectrum and doppler shift features by nonlinear transform in our identification method, while maintaining a balance between local and global feature extraction. To evaluate the model's performance, we conducted an extensive data collection campaign, acquiring signals from Starlink satellites over a 25-day period, resulting in a dataset comprising over 30,000 data samples. Experimental results demonstrate that our proposed method achieves superior accuracy (91.3%) compared to existing approaches in the field of satellite signal identification.

Information leakage through wireless channels poses a significant security concern within contemporary cellular networks, such as 5G new radio (NR). Among the myriad of potential attack vectors, passive traffic analysis (PTA) stands out as a pervasive and surreptitious threat, which allows attackers to discern the specific services utilized by unsuspecting victims without their noticing. In this work, we present a pioneering approach to achieve fine-grained service identification by adopting an unexplored perspective: mapping traffic transmission patterns to physical layer time-frequency occupancy patterns, which we refer to as Passive Time-Frequency Traffic (PTTF). Additionally, it selects the uplink control channel that carries the acknowledgment/negative acknowledgment (ACK/NACK) feedback within the Hybrid Automatic Repeat reQuest (HARQ) process as the data source. Statistical features of ACK/NACK time-frequency resources are extracted for traffic classification, and activities are recognized from a three-tier classification algorithm. For validation, we conduct field experiments targeting commercialized smartphones within the real-world operator’s network. This setup effectively mirrors practical scenarios, as the resources within the target frequency band can also be allocated to other equipment in the operator’s network. Furthermore, cross-validation experiments involving different smartphone brands and various network formats are conducted in order to ascertain the generalizability of the proposed PTTF.

With the explosive growth in the number of NSGO satellites, there is an increased demand for satellite tracking technology, making the accurate prediction of satellite trajectory and optimization of target tracking efficiency crucial. This paper proposes a strategy for tracking NGSO satellites, which employs a two-stage stochastic model with a greedy approach to optimize the real-time prediction of multiple NGSO satellite passes and improve the efficiency of noncooperative ground equipment. The first stage of the model employs SGP4/SDP4 functions for initial orbit predictions, isolating satellite data from ground equipment performance. The second stage iteratively refines these predictions, addressing the complexities associated with noncooperative ground equipment, constrained resources, the high frequency of NGSO constellation satellites pass, and effective target management. The resulting two-stage model achieves precise and efficient operations, effectively overcoming the traditional challenges of stochastic optimization in dynamic conditions involving multiple satellites. It successfully copes with the exponential surge in computational requirements as the number of satellites increases. A case study with an antenna system validates the practicality and effectiveness of the model. The optimized sequence significantly enhances overall performance concerning efficiency, utilization, idle time, angular displacement, and trajectory symmetry, achieving a 27.48% improvement over the traditional solution and a notable 58.33% increase in the number of effectively tracked satellites, all within a rapid 15-second execution window.

Malicious traffic detection is an important task in network security, which protects the target network from privacy leakage and service paralysis. The complexity of the network and the hierarchical structure of network traffic, i.e, byte-packet-flow, indicate the diversity of traffic information. Most of the existing work only uses one feature or statistical feature, and cannot learn network traffic from multiple perspectives, i.e, shortsighted, which results in the lack of important information in network traffic. Meanwhile, after obtaining multiple features, the effective fusion of multiple features is also an urgent problem to be solved. In this paper, we propose a Multiple Features Fusion with Attention Networks (MFFAN). According to the hierarchical structure of network traffic, we extract byte, packet, and statistical features from original traffic files to learn traffic from multiple perspectives, overcoming shortsighted. To effectively fuse multiple features, we use the self-attention to learn the intra-feature relationship with each feature and use the co-attention to learn the inter-feature relationship between features. We conduct experiments on the ISCIDS2012 dataset and CICIDS2017 dataset, and the results show that our model achieves an effective fusion of multiple features and high accuracy.

Automatic modulation classification (AMC), which aims to identify the modulation types of unknown signals without any prior knowledge, plays a key role in intelligent wireless communication. In recent years, the outstanding achievements of deep learning in the fields of computer vision and nature language processing have promoted the continuous researches of deep learning in AMC. In view of the fact that the existing deep learning-based AMC ignores the frequency attribute of the modulated signals, this paper proposed an AMC approach based on deep learning which comprehensively considers the contribution of frequency and spatiotemporal characteristics of the modulated signals. Specifically, we utilize the sliding window to divide the raw signals into short-time signal slices. In order to obtain the representation of frequency domain of the signals, each short-time signal slice is decomposed into several modes by variational mode decomposition (VMD), which is an adaptive signal decomposition technique. Finally, all the modes are used as input of the convolutional neural network (CNN) as a tensor to learn high-level features from the frequency and spatiotemporal domain to identify the modulation type. In this paper, we verify the effectiveness of the proposed approach with 10 kinds of modulation types generated by SMW200A Vector Signal Generator. More importantly, the proposed approach achieves 99.2% of the overall classification accuracy at 8dB signal to noise ratio (SNR) and shows better robustness to noise than the existing AMC approaches.

Automatic wireless spectrum anomaly detection is vital to intelligent management of electromagnetic spectrum, which aims to detect various jamming and anomalous working states, especially intentional jamming. The intentional jamming has evolved in a variety of ways, but the existing spectrum anomaly detection efforts give little consideration to the diverse intentional jamming. Here, we firstly generate a rich dataset consisting of five types of normal signals and four types of intentional jamming. In order to effectively detect anomalies, we propose AWGAN, a novel anomaly detection method based on Wasserstein generative adversarial network. AWGAN can not only learn the distribution of normal time-frequency waterfall images in a latent space, but also remember the detailed features of normal images, and generate same images as the normal images by adversarial training. To detect anomalies, we propose a random reverse mapping (RRM) method based on backpropagation, to map a new time-frequency waterfall image into the latent space, so as to find the vector closest to the distribution of the new image in the latent space. We also define a scoring criterion to score images indicating their fit into the learned distribution. The experimental results show that the comprehensive detection ability of our method is superior to other methods for detecting the four types of anomalies.

This study attempts to characterize the electromagnetic compromising emanations from Laser Facsimile. Electromagnetic radiation is inevitable when electronic equipment works. The radiations can deteriorate the performance of a part of device itself or of another system. This phenomenon is a subject in the field of electromagnetic compatibility (EMC). But concerning people who process classified information, there is another threat that such electromagnetic radiation could be intercepted, deciphered and finally reveal the data being processed by the equipment. In this paper, electromagnetic side channel of laser facsimile and countermeasures are analyzed and discussed. It shows that the electromagnetic side channel of electronic information equipments is a real threat of information security.

Computer displays emit electromagnetic waves, which compromise the information displayed by the computer. This can be a potential information security threat as the sensitive information can be stolen from a distance without leaving any trace. The video leakage signals contain the information of the image displayed in the computer, so the video leakage signals can be seen as special image signals. However, different from the normal image signal, the signal to noise ratio (SNR) of video leakage signal is low due to the environmental noise and many other man-made noises. In this paper, a novel wavelet based independent component analysis (ICA) method is proposed for improving SNR of computer video leakage signals. By using this method, we can improve the performance of pre-processing of computer video leaking signals. We solve the problem of using Fast ICA in processing video leakage signal by using a wavelet filter. The performance of Fast ICA is improved by working in wavelet domain because of advantages like ease of implementation and less computation time when compared to time domain. We pre-process one-dimensional received signal without reconstructing the image since it is inefficient to reconstruct signal before processing. A direct SNR can't be defined. Therefore, another metric called quasi signal to noise ratio (QSNR) is defined to estimate signal to noise ratio of video leakage signals. The processed results of the actual experimental data show that the proposed wavelet based ICA algorithm has a better performance than the Fast ICA algorithm and the Wavelet denoising algorithm.

Modulation recognition has been an essential part in non-cooperative communications. Without knowing any prior knowledge of the transmitted data, the received signals may be affected by intersymbol interference (ISI), carrier frequency offset, phase offset and timing delay. ISI and these unknown parameters would degrade the performance of many recognition methods. There have been many robust estimators of carrier frequency offset, phase offset and timing delay. However, the robust recognition methods combating ISI have not been well designed. Most blind channel equalization methods, which are designed to overcome the effects of ISI, have poor robustness to noise and carrier frequency offset. In this paper, a robust recognition scheme is proposed to recognize phase-shift keying (PSK) and quadrature amplitude modulation (QAM) signals in non-cooperative communication with ISI environment. The envelope spectrum of PSK and QAM signals are analyzed. Then the recognition scheme based on the ratio of spectral peak to spectral centroid (RSPSC) of envelope spectrum is presented. Such scheme achieves a good recognition performance in the presence of ISI. Spectral centroid is utilized to enhance the robustness to noises. By extracting the envelope spectrum, in addition, the method avoids being affected by carrier frequency offset, phase offset and timing delay. The feasibility of the proposed method has been proved in the respects of theory and simulations.

Considering that the text information might be leaked through electromagnetic radiation from a computer display, a novel algorithm has been developed to detect the text information leakage. Motivated by the observation that the ‘text ‐ space – text’ characteristic for electromagnetic radiation signal contains the text information, the authors proposed a method to describe this characteristic. In this method, sparse decomposition in wavelet was used and sub‐band sparsity was defined. Variance mean ratio and correlation coefficients of sub‐band sparsity were combined as the features of electromagnetic radiation signals. By using this method, the authors can accurately and efficiently detect the text information leakage in electromagnetic radiation from a computer display without reconstructing the displayed image.

Electromagnetic radiation signal from computer display can be seen as a computer security risk if the radiation signal is intercepted and reconstructed. Electromagnetic radiation signal from computer display can also be called video leaking signal. Synchronising information extraction is the key problem of computer video leaking signal interception and reconstruction. To solve such problem, a novel synchronising information extraction algorithm based on spectral centroid has been developed. This study not only introduced spectral centroid into video leaking signal processing but also defined the concept of segmented spectral centroid. In addition, the uniformity degree of spectral centroid spacing distribution was defined to describe the harmonic characteristics of video leaking signal spectrum. The proposed algorithm can extract the electromagnetic radiation signal's synchronising information automatically and efficiently even with interference signal. Thus, the interception and reconstruction of electromagnetic radiation can be realised more effectively and the anti‐interference performance can be improved.

A Polarized OFDM (POFDM) scheme is proposed for the wireless depolarized channel. Through such scheme, each sub-carrier's polarization state, amplitude and phase can be modulated to bear information, and the spectrum efficiency can be further improved. Consider the multi-path fading channel, the channel caused depolarization impairments to POFDM are quantitatively analyzed; to mitigate such impairments, a polarization constellation compensation algorithms is also presented. Furthermore, based on the proposed compensation algorithm, the POFDM's symbol error rate is derived, then the link spectrum efficiency optimization model is established. Through calculation, the optimal modulation ratio is found to maximize POFDM's spectrum efficiency. Finally, both numerical simulations and experimental results show that POFDM can significantly improve the link spectrum efficiency compared with the traditional OFDM scheme.