Identifying indoor individuals using micro-Doppler signature of multiple-input multiple-output (MIMO) through-the-wall radar (TWR), and determining whether they pose a threat holds significant research value in the field of urban security surveillance. However, large-scale TWR human motion data is difficult to collect, which reduces the recognition performance. To address this issue, a MIMO TWR micro-Doppler signature representation method under limited data based on heterogeneous transfer learning is proposed in this letter. The multi-channel TWR human motion Doppler-time maps (DTMs) are first generated, and the trace-ratio group sparse method is then proposed for multi-channel DTM feature augmentation. In addition, a micro-Doppler signature representation method based on optimal transport domain adaptation heterogeneous transfer learning is proposed. By leveraging large-scale millimeter-wave radar human gait data, the proposed method guides the micro-Doppler signature representation to maximize inter-class separation on the TWR DTM set. The effectiveness of the proposed method is validated through a few-shot measured dataset collected for TWR human threat identification.

In GPS-denied scenarios, a robust environmental perception and localization system becomes crucial for autonomous driving. In this paper, a LiDAR-based online localization system is developed, incorporating road marking detection and registration on a high-definition (HD) map. Within our system, a road marking detection approach is proposed with realtime performance, in which an adaptive segmentation technique is first introduced to isolate high-reflectance points correlated with road markings, enhancing real-time efficiency. Then, a spatio-temporal probabilistic local map is formed by aggregating historical LiDAR scans, providing a dense point cloud. Finally, a LiDAR bird’s-eye view (LiBEV) image is generated, and an instance segmentation network is applied to accurately label the road markings. For road marking registration, a semantic generalized iterative closest point (SG-ICP) algorithm is designed. Linear road markings are modeled as 1-manifolds embedded in 2D space, mitigating the influence of constraints along the linear direction, addressing the under-constrained problem and achieving a lower localization errors on HD maps than ICP. Extensive experiments are conducted in real-world scenarios, demonstrating the effectiveness and robustness of our system.

Transductive Few-Shot Learning (TFSL) has recently attracted increasing attention since it typically outperforms its inductive peer by leveraging statistics of query samples. However, previous TFSL methods usually encode uniform prior that all the classes within query samples are equally likely, which is biased in imbalanced TFSL and causes severe performance degradation. Given this pivotal issue, in this work, we propose a novel Conditional Transport (CT) based imbalanced TFSL model called Prototypes-oriented Unbiased Transfer Model (PUTM) to fully exploit unbiased statistics of imbalanced query samples, which employs forward and backward navigators as transport matrices to balance the prior of query samples per class between uniform and adaptive data-driven distributions. For efficiently transferring statistics learned by CT, we further derive a closed form solution to refine prototypes based on MAP given the learned navigators. The above two steps of discovering and transferring unbiased statistics follow an iterative manner, formulating our EM-based solver. Experimental results on four standard benchmarks including miniImageNet, tiered-ImageNet, CUB, and CIFAR-FS demonstrate superiority of our model in class-imbalanced generalization1.

Infection is a fault attack countermeasure, which aims to destroy the dependency of the faulty ciphertexts on the secret key. However, current security evaluations on infection countermeasures are either tailored for the specific attack scenario or not general enough to apply to various infection instances. They cannot come to convincing results, let alone make comparisons between different countermeasures. Based on information theory, this paper presents a generic evaluation framework that is feasible for various infection countermeasures and attack scenarios. The framework is constructed with the idea to separate the infection function from the unprotected cipher yet consider the fault injection effect on the unprotected cipher in the infection function evaluation. First, the security judging criteria for the infection function under different attack scenarios are personalized according to the injection-caused security loss of the unprotected cipher. Then, a universal method of security quantitative analysis on infection function is proposed with two important steps: the prior knowledge collection and the infection operation decomposition analysis. Because the analysis results of the simple infection operation can be reused within the infection function under various attack scenarios, the security quantifications are efficient. Based on this framework, the paper also reviews some existing infection countermeasures for their fault attack resistances. The result shows that our analysis can expose more infection vulnerabilities than the previous works. Besides, the security quantification and judgment on these countermeasures give us a new insight into their security applicable scopes. They are instructive for the countermeasure selection when the implementation costs are very close. Furthermore, the framework provides an efficient way to evaluate future infection countermeasures.

SIMON is a new family of lightweight block ciphers proposed by the National Security Agency (NSA) in 2013. Since its publication, it has attracted much research interest and a number of analysis results have been presented. As a popular kind of implementation attack method, the fault attack also works when it is applied to SIMON. In this paper, we propose an effective fault attack on SIMON under the random byte fault model. Compared with the previous attack results, our attack can successfully recover the whole master key with injecting the faults into only one intermediate round for six instances of SIMON. In our attack, we fully utilize a class of differential propagation properties of SIMON to determine the fault injection position as long as the full diffusion of the fault has not been obtained. On the basis of it, we can recover the last round key with the differential analysis technique. The differential propagation properties make it possible to inject the faults into the earlier intermediate round at the beginning than that of the previous attacks. Meanwhile, the same faulty ciphertext set can also help to recover other round keys. So we do not have to inject the faults into any other intermediate rounds to reveal the whole master key. Moreover, in this paper we also give a detailed mathematical analysis on the average number of the fault injections under the random byte fault model. The data complexity analysis shows that less fault injections are required in our attack compared with other work under the same attack model. Finally, we also verify the effectiveness and correctness of our attack with experiments.