Kamruzzaman Sarker

In the local contexts such as distributed internet of things (IoT)-based monitoring or control systems, the impact of cybersecurity attacks is amplified by limited computational resources and the lack of adaptive encryption mechanisms. To address these challenges, this study proposes a lightweight and chaos-driven encryption model that integrates elliptic curve cryptography (ECC) with dynamic key generation based on logistic map perturbation and adaptive bit-level rotation. The proposed method is designed to enhance payload integrity while minimizing computational overhead. Experimental results demonstrate that the model achieves low latency with an average encryption time of 4.97 ms and decryption time of 6.28 ms. Also, it exhibits a payload overhead of only 18% and maintains 100% correctness under targeted packet injection scenarios. These results confirm the suitability of the model for resource-constrained environments. This work contributes a novel symmetric encryption approach with practical robustness and scalability. It paves the way for secure-by-design network protocols in real-time applications.

Recent advances in artificial intelligence (AI) and especially in deep learning have manifested an increasing concern in trustworthiness. Neuro-symbolic methods, which mix some elements of neural networks with some elements of symbolic reasoning, have shown great potential for some aspects of trustworthiness, particularly for interpretability. In this article, we provide an overview of the various ways neuro-symbolic methods have been used to increase the trustworthiness, in the latest literature of the leading conferences. In particular, we focus on the contributions of the recent articles that achieve better interpretability thanks to NeSy systems, while also considering contributions in a broader sense, such as safety, fairness, and privacy. We also did a categorization of the existing contributions along several key dimensions related to the symbolic structures they are exploiting, and the type of interpretability they provide.

With the widespread adoption of deep learning techniques, the need for explainability and trustworthiness is increasingly critical, especially in safety-sensitive applications and for improved debugging, given the black-box nature of these models. The explainable AI (XAI) literature offers various helpful techniques; however, many approaches use a secondary deep learning-based model to explain the primary model’s decisions or require domain expertise to interpret the explanations. A relatively new approach involves explaining models using high-level, human-understandable concepts. While these methods have proven effective, an intriguing area of exploration lies in using a white-box technique to explain the probing model. We present a novel, model-agnostic, post hoc XAI method that provides meaningful interpretations for hidden neuron activations. Our approach leverages a Wikipedia-derived concept hierarchy, encompassing approximately 2 million classes as background knowledge, and uses deductive reasoning-based concept induction to generate explanations. Our method demonstrates competitive performance across various evaluation metrics, including statistical evaluation, concept activation analysis, and benchmarking against contemporary methods. Additionally, a specialized study with large language models (LLMs) highlights how LLMs can serve as explainers in a manner similar to our method, showing comparable performance with some trade-offs. Furthermore, we have developed a tool called ConceptLens, enabling users to test custom images and obtain explanations for model decisions. Finally, we introduce an entirely reproducible, end-to-end system that simplifies the process of replicating our system and results.

This study presents a multi-level image encryption framework that combines password dependent SHA-256 key generation with a Logistic Map-based chaotic mechanism, supporting three operational modes: Speed, Balanced, and Security. The system is designed for scalability and robustness across diverse image sizes, achieving up to 27 percent faster encryption than AES on 1024×1024 images while maintaining high cryptographic strength. Experimental results show strong randomness with entropy reaching up to 7.98 bits per pixel, reduced adjacent pixel correlation below 0.01, and high resistance to differential attacks with NPCR above 99.6 percent and UACI around 33.4 percent. Structural integrity after decryption is also preserved with SSIM scores above 0.98. Compared to existing chaos based methods such as those proposed by Arif et al. and Riaz et al., the proposed system offers superior entropy performance, enhanced flexibility through multi-mode encryption, and broader resolution support up to 2048×2048 pixels. Comprehensive evaluations using entropy, correlation, PSNR, SSIM, XOR, and 3D histogram analysis confirm the method’s effectiveness. These findings highlight the system’s suitability for real-time, secure image transmission in environments such as IoT, medical imaging, and embedded applications.

Digital image security is increasingly vulnerable to sophisticated attacks, underscoring the urgent need for robust encryption techniques. Traditional encryption methods often fall short in defending against advanced threats, highlighting the importance of innovative solutions to protect digital images. This study tackles these challenges by incorporating quantum computing into image encryption, employing techniques such as bit-plane scrambling, pixel permutation, and bit permutation. These strategies enhance security by introducing complex, non-linear transformations that make decryption attempts significantly more difficult without the correct cryptographic keys. A key configuration based on r=44, μ=2024 is employed to achieve this. The integration of quantum bit-plane scrambling and quantum pixel permutation results in a highly secure encryption method. Experimental results show substantial improvements in entropy levels, along with strong unified average changing intensity (UACI) and number of pixels change rate(NPCR) values across various images. Notably, the "Peppers" image achieved the best performance, with UACI values of 33.5572 and NPCR values of 99.8301. The method proves highly effective, as repeated tests with incorrect keys failed to decrypt the plain image accurately. Future research could explore the addition of a discrete quantum wavelet transform to further enhance the security and efficiency of quantum-based image encryption methods.