Images, as carriers of rich information, are generated, stored, and transmitted in various forms across diverse scenarios. It has become an important issue in the field of information security today to encrypt images to ensure information security. To address this issue, this paper proposes a Pixel-Level and DNA-Level Image Encryption Method Based on a Five-Dimensional Hyperchaotic System, named PD5H. The proposed method combines a five-dimensional chaotic system, a novel pixel-block internal diffusion method, and a new flow diffusion method integrating Pixel-Level and DNA-Level encryption, hereinafter referred to as ‘joint diffusion’. The improved 5D chaotic system can generate highly complex and unpredictable chaotic sequences. The intra-block diffusion process utilizes the internal information of the image to perform preliminary diffusion and reduce pixel correlation. The joint diffusion process can effectively employ various encryption methods to encrypt images with different step sizes at the bit level. PD5H has a large key space, extremely low image correlation, a uniform ciphertext pixel distribution, an excellent ciphertext entropy value (>7.999), and strong resistance to differential attacks. It also demonstrates strong resistance to data loss. The security analysis confirms that PD5H demonstrates excellent performance in color image encryption and can effectively resist various common attacks.

While Neural Architecture Search (NAS) has revolutionized the automation of deep learning model design, gradient-based approaches like DARTS often suffer from high computational overheads, the collapse of skip-connections, and optimization instability. To address these limitations, we propose Efficient and Lightweight Differentiable Architecture Search (EL-DARTS). EL-DARTS constructs a compact and redundancy-reduced search space, integrates a partial channel strategy to lower memory usage, employs a Dynamic Coefficient Scheduling Strategy to balance edge importance, and introduces entropy regularization to sharpen operator selection. Experiments on CIFAR-10 and ImageNet demonstrate that EL-DARTS substantially improves both search efficiency and accuracy. Remarkably, it attains a 2.47% error rate on CIFAR-10, requiring merely 0.075 GPU-days for the search. On ImageNet, the discovered architecture achieves a 26.2% top-1 error while strictly adhering to the mobile setting (<600 M MACs). These findings confirm that EL-DARTS effectively stabilizes the search process and pushes the efficiency frontier of differentiable NAS.