To address the problem of low detection rate caused by the close alignment and multi-directional position of text words in practical application and the need to improve the detection speed of the algorithm, this paper proposes a multi-directional text detection algorithm based on improved YOLOv3, and applies it to natural text detection. To detect text in multiple directions, this paper introduces a method of box definition based on sliding vertices. Then, a new rotating box loss function MD-Closs based on CIOU is proposed to improve the detection accuracy. In addition, a step-by-step NMS method is used to further reduce the amount of calculation. Experimental results show that on the ICDAR 2015 data set, the accuracy rate is 86.2%, the recall rate is 81.9%, and the timeliness is 21.3 fps, which shows that the proposed algorithm has a good detection effect on text detection in natural scenes.

An improved DETR (detection with transformers) object detection framework is proposed to realize accurate detection and recognition of characters on shipping containers. ResneSt is used as a backbone network with split attention to extract features of different dimensions by multi-channel weight convolution operation, thus increasing the overall feature acquisition ability of the backbone. In addition, multi-scale location encoding is introduced on the basis of the original sinusoidal position encoding model, improving the sensitivity of input position information for the transformer structure. Compared with the original DETR framework, our model has higher confidence regarding accurate detection, with detection accuracy being improved by 2.6%. In a test of character detection and recognition with a self-built dataset, the overall accuracy can reach 98.6%, which meets the requirements of logistics information identification acquisition.

The return-oriented programming (ROP) is a prevalent code reuse attack. It exploits vulnerabilities such as stack overflows to hijack the execution flow into the carefully chosen instruction snippet (called gadget) chain already in the memory. Address Space Layout Randomization (ASLR) is widely used to defend against code reuse attacks like ROP from hijacking control flow. However, the Linux kernel, especially the most vulnerable kernel module, only applies basic coarsegrained load-time ASLR methods. This is far from sufficient to defend against new ROP attacks. In this paper, we propose Chaos, a fine-grained runtime architecture-independent ASLR for the kernel module. Chaos adopts function granularity randomization to increase randomization entropy. It continuously performs randomization during runtime to invalidate memory disclosure. To reduce overhead, Chaos only suspends the kernel module and minimizes the running state migration efforts. We implement Chaos based on Linux 5.10 LTS on ARM64 and evaluate it using two common benchmarks. The result demonstrates that Chaos achieves fine-grained runtime ASLR for protecting kernel modules with 5% performance overhead degradation at most.

Learning unlabeled samples from unseen categories given limited labeled data is a challenging problem. Existing few-shot learning methods fail to generate satisfactory feature representations due to tackling the informative and interference information without distinction. In this paper, we propose an attention-guided two-stream convolutional neural network (AGTSNet), which highlights the salient and discriminative features of the main object while alleviating the background interference to address this indiscriminate treatment. Comprehensive experiments are conducted on few-shot image classification with four standard benchmark datasets to demonstrate the effectiveness of our method.