Data security is critical, particularly in medical imaging, yet remains challenging. Many research efforts have focused on enhancing medical image security, particularly during network transmission. Ensuring confidentiality and authenticity is a key priority for researchers. However, traditional encryption methods are unsuitable for IoT environments due to data size limitations. Lightweight encryption algorithms that preserve confidentiality, integrity, and authenticity can address these limitations. This paper proposes an efficient, lightweight method to encrypt and authenticate medical images in healthcare systems. The approach splits images into diagonal and non-diagonal blocks, and then processes them in two phases: (1) non-diagonal blocks are permuted using inter-block differences and XORed with diagonal blocks for substitution; (2) diagonal blocks are encrypted via AES and enhanced CBC mode with a tag mechanism for integrity. Security tests (histograms, correlation, entropy, NPCR, UACI) verify the scheme’s robustness. The results show that the model outperforms existing techniques in efficacy and attack resistance, making it viable for medical IoT and smart surveillance.

The increasing deployment of Unmanned Aerial Vehicles (UAVs) in critical operations exposes them to cyberattacks. Although deep learning-based Intrusion Detection Systems (IDSs) are effective, they are susceptible to adversarial attacks that manipulate input features to avoid detection. Conventional feature selection methods do not distinguish between features critical to model accuracy and those essential for preserving cyberattack operational validity. To address this, we propose a Dual-Impact Feature Selection (DIFS) framework for robust UAV-IDS models. Our approach evaluates features based on two criteria: the first is Model Performance Impact (MPI), using Integrated Gradients (IG) and Local Interpretable Model-agnostic Explanations (LIME) to measure feature influence on detection accuracy, and the second is Functionality Preservation Criterion (FPC), a clustering-based method that assesses whether a feature is indispensable for cyberattack execution. Features with high MPI and FPC are identified as Dual-Impact Features (DIFs). We generate constrained adversarial attacks that perturb these DIFs to create realistic evasion samples. Using these samples for adversarial training, we develop three robust UAV-IDS Convolutional Neural Network (CNN) models. Evaluated on three UAV network intrusion datasets, our framework demonstrates improved resilience. The models achieve up to 99.8% detection accuracy while reducing Attack Success Rate (ASR) to as low as 0.002, supporting their potential for designing adversary-resistant detection systems for UAV networks.

The rapid growth of the Internet of Things (IoT) and edge computing has increased the demand for secure and lightweight image encryption suitable for resource-constrained environments. This paper proposes a hybrid framework combining a residual-based pretrained autoencoder with a novel Cylinder XOR-Cascade (CXC) encryption scheme. The autoencoder compresses images into a compact latent representation while a residual branch preserves fine spatial details for accurate reconstruction. Both representations are encrypted using CXC, a two-pass column-wise stream cipher that enhances confusion and diffusion through sequential SHA3-256-based chaining and a cylinder-like feedback mechanism. Experiments on the USC-SIPI dataset demonstrate strong statistical security: Number of Pixels Change Rate (NPCR) ≈99.5%, Unified Average Changing Intensity (UACI) ≈33.5%, and entropy ≈7.5. The full pipeline processes images in 280 ms on GPU and 928 ms on CPU, with the cryptographic stage requiring only 25 ms. For noisy channels, integrating Reed–Solomon forward error correction restores Structural Similarity Index Measure (SSIM) >0.85 under Gaussian noise (σ=20), making the framework suitable for edge-assisted IoT and real-time visual transmission across both reliable and noisy channel conditions.