ABSTRACT As network throughput increases and security threats escalate in big‐data environments, shallow machine‐learning models are inadequate for handling large‐scale network traffic. Aiming at the common class imbalance and high dimension problem of intrusion detection data sets, this paper proposes SSA‐CL (Sparrow Search Algorithm‐CNN‐LSTM) model based on SSA (Sparrow Search Algorithm) and CNN‐LSTM (Convolutional Neural Network, Long Short‐Term Memory) model. SSA‐CL model uses CNN and LSTM to build a CNN‐LSTM hybrid model that can automatically extract features and effectively process sequence data. We apply SMOTE to mitigate class imbalance and employ SSA to optimize hyperparameters, significantly improving detection accuracy and efficiency. The experimental results show that the proposed method has significant advantages in key indicators such as accuracy and recall, with excellent multi‐classification effect. SSA‐CL combines a compact architecture, fast training, and strong detection performance, indicating practical value for network‐security applications.

Real-time intrusion detection with millisecond response is critical for Internet of Vehicles (IoV) security but is challenged by extreme class imbalance and high computational costs. This paper proposes a novel multimodal framework integrating Denoising Diffusion Probabilistic Models (DDPM) and Knowledge Distillation (KD). First, multi-source data is transformed into RGB images. A conditional DDPM with timestep and class embeddings balances datasets by generating minority-class samples. The teacher model (DiffuGuardian) fuses text-image features for training. Subsequently, a lightweight student model, LiteSentinel, is designed employing depthwise separable convolutions and inverted residual blocks to reduce parameters. Results on three datasets demonstrate that DiffuGuardian consistently achieves around 98–100% precision, accuracy, recall, and F1-score under 5-fold evaluation, while LiteSentinel maintains approximately 95–99% across all metrics with substantially reduced complexity. DiffuGuardian reaches an inference time of 3.80ms with a model size of 0.10 MB, whereas LiteSentinel further reduces latency to 0.79ms with a size of 0.07 MB, enabling efficient edge deployment for IoV security.

With the rapid growth of network traffic and the widespread adoption of encryption, traditional detection methods have increasingly revealed limitations in feature extraction. To address this issue, this paper proposes a detection model called the Graph Convolutional Network–Gated Recurrent Unit–Transformer Network (GGT-Net). The model represents network traffic separately as a protocol graph structure and a statistical sequence: a Graph Convolutional Network (GCN) extracts TLS protocol interaction information, a Gated Recurrent Unit (GRU) captures statistical behaviors, and a Transformer deeply fuses the two modalities. Skip connections are introduced to enhance the expressive power of the GCN, while contrastive loss is employed to strengthen the consistency of multimodal representations. Experimental results demonstrate that GGT-Net achieves significantly better accuracy and recall than existing methods, and also exhibits strong generalization across multiple benchmark datasets.

To alleviate the contradiction between image privacy in Internet of Things (IoT) cloud storage platforms and the convenience of platform functionality, a class of visual image privacy preservation schemes, Thumbnail Preserving Encryption (TPE), is proposed. These schemes achieve a balance between preserving image privacy and usability by altering the size of the thumbnail block, which makes the encrypted image well compatible with the convenient features of the IoT cloud storage platform. However, these schemes are flawed in that the color vividness of the encrypted image remains constant regardless of circumstances. These schemes may not be suitable for various applications in different Industrial IoT scenarios since the intensity of color can directly influence individuals’ understanding and assessment of image semantics. To this end, a TPE scheme with gradable color vividness is proposed, which achieves the gradual change of color vividness of encrypted images from dull or distorted to bright or real under any thumbnail block size. In this scheme, the highest bit plane of the image is isolated for block space adjustment, while the remaining 7-bit planes are encrypted using substitution-permutation encryption. By adjusting the sum of the elemental values in the block space to gradually approximate the sum of the elemental values in the original block pixels, the encrypted image and its thumbnail create the effect of a gradual change in color vividness. Furthermore, the vibrancy of the color intensifies as the adjustment ratio is increased. The experimental results show that the scheme does not significantly affect the privacy of the encrypted images, despite the gradual change in color vividness.

Images are generally uploaded to the cloud in plaintext and can be retrieved in the cloud, but privacy may be exposed. To solve this problem, Privacy Preserving Content Based Image Retrieval (PPCBIR) system was proposed. In this system, noise-like image encryption algorithm was used in the early scheme, and Thumbnail Preserving Encryption (TPE) technology was proposed later to balance image privacy and visual usability. However, the existing TPE schemes supporting retrieval have shortcomings in mining the visual usability of TPE images, which limits the retrieval accuracy. Based on this, we propose a VF-PPCBIR scheme combining TPE and image visual features to improve retrieval efficiency and accuracy while ensuring image privacy. Specifically, we redesign a new TPE algorithm for lossless encryption and decryption of arbitrary size images. The design concept of the encryption algorithm is novel, and the encryption effect is more stable. The retrieval process generates thumbnails of the retrieved image and extracts local features in the spatial domain, which are matched with the features extracted from TPE thumbnails in the cloud, and the user can directly select the desired image. In addition, the retrieval scheme uses adjustable feature algorithm to achieve approximate similarity between the ciphertext and the plaintext thumbnail, to achieve accurate feature matching. The experimental results show that the time cost, and mean average precision (mAP) can reach 9.121s and 64.343%, respectively.

In traditional image encryption, privacy is protected at the expense of all visual content resulting in poor usability. Recently, a novel image encryption concept, thumbnail-preserving encryption (TPE), has been proposed to balance privacy and visual usability after encryption. However, the existing TPE schemes can only encrypt images with specific sizes (related to thumbnail block sizes). The thumbnail block, namely, the image sub-block of equal length and width, and specific means the image size can be divisible by the thumbnail block size. In fact, a little thought reveals that in reality the image size is arbitrary, and it is only by chance that images can be encrypted fully. To this end, we propose a generalized TPE scheme, and it realizes full encryption of images with arbitrary size. Specifically, first of all, a novel block technique called size-compatible-match is proposed. It can be used to accurately match and segment the portion of the image that cannot be encrypted by existing TPE schemes. Secondly, a chaotic system called 2D-GMOS is introduced to greatly reduce the time cost of the encryption and decryption process. Third, the block technique and 2D GM-OS chaotic system are combined with the TPE. The results have demonstrated that images with arbitrary size can be fully encrypted (no leakage of the original image) by the proposed scheme, and the encrypted image has useful visual meaning. Meanwhile, extensive experiments have been done that show the security and robustness of the proposed scheme.

Linear secret-sharing scheme (LSSS) is a useful tool for supporting flexible access policy in building attribute-based encryption (ABE) schemes. But in lattice-based ABE constructions, there is a subtle security problem in the sense that careless usage of LSSS-based secret sharing over vectors would lead to the leakage of the master secret key. In this paper, we propose a new method that employs LSSS to build lattice-based key-policy attribute-based encryption (KP-ABE) that resolves this security issue. More specifically, no adversary can reconstruct the master secret key since we introduce a new trapdoor generation algorithm to generate a strong trapdoor (instead of a lattice basis), that is, the master secret key, and remove the dependency of the master secret key on the total number of system attributes. Meanwhile, with the purpose of reducing the storage cost and support dynamic updating on attributes, we extended the traditional 1-dimensional attribute structure to 2-dimensional one. This makes our construction remarkably efficient in space cost, with acceptable time cost. Finally, our scheme is proved to be secure in the standard model.

Most anti-collusion audio fingerprinting schemes are aiming at finding colluders from the illegal redistributed audio copies. However, the loss caused by the redistributed versions is inevitable. In this letter, a novel fingerprinting scheme is proposed to eliminate the motivation of collusion attack. The audio signal is transformed to the frequency domain by the Fourier transform, and the coefficients in frequency domain are reversed in different degrees according to the fingerprint sequence. Different from other fingerprinting schemes, the coefficients of the host media are excessively modified by the proposed method in order to reduce the quality of the colluded version significantly, but the imperceptibility is well preserved. Experiments show that the colluded audio cannot be reused because of the poor quality. In addition, the proposed method can also resist other common attacks. Various kinds of copyright risks and losses caused by the illegal redistribution are effectively avoided, which is significant for protecting the copyright of audio.

As a subset of discrete Fourier transform (DFT), discrete cosine transform (DCT), especially two-dimensional discrete cosine transform (2D-DCT), is an important mathematical tool for digital signal processing. However, the computational complexity of 2D-DCT is quite high, which makes it impossible to meet the requirements in some signal processing fields with large signal sizes. In addition, to optimize the 2D-DCT algorithm itself, seeking help from a cloud platform is considered to be an excellent alternative to dramatically speeding up 2D-DCT operations. Still, there are three key challenges in cloud computing outsourcing that need to be addressed, including protecting the privacy of input and output data, ensuring the correctness of the returned results, and ensuring adequate local cost savings. In this paper, we explore the design of a practical outsourcing protocol for 2D-DCT and 2D-IDCT, which well solves the above three challenges. Both theoretical analysis and simulation experiment results not only confirm the feasibility of the proposed protocol but also show its outstanding performance in efficiency.