The discipline of automatic image captioning represents an integration of two pivotal branches of artificial intelligence, namely computer vision (CV) and natural language processing (NLP). The principal functionality of this technology lies in transmuting the extracted visual features into semantic information of a higher order. The bidirectional long short-term memory (Bi-LSTM) has garnered wide acceptance in executing image captioning tasks. Of late, scholarly attention has been focused on modifying suitable models for innovative and precise subtitle captions, although tuning the parameters of the model does not invariably yield optimal outcomes. Given this, the current research proposes a model that effectively employs the bidirectional LSTM and attention mechanism (Bi-LS-AttM) for image captioning endeavors. This model exploits the contextual comprehension from both anterior and posterior aspects of the input data, synergistically with the attention mechanism, thereby augmenting the precision of visual language interpretation. The distinctiveness of this research is embodied in its incorporation of Bi-LSTM and the attention mechanism to engender sentences that are both structurally innovative and accurately reflective of the image content. To enhance temporal efficiency and accuracy, this study substitutes convolutional neural networks (CNNs) with fast region-based convolutional networks (Fast RCNNs). Additionally, it refines the process of generation and evaluation of common space, thus fostering improved efficiency. Our model was tested for its performance on Flickr30k and MSCOCO datasets (80 object categories). Comparative analyses of performance metrics reveal that our model, leveraging the Bi-LS-AttM, surpasses unidirectional and Bi-LSTM models. When applied to caption generation and image-sentence retrieval tasks, our model manifests time economies of approximately 36.5% and 26.3% vis-a-vis the Bi-LSTM model and the deep Bi-LSTM model, respectively.
To protect image privacy in real-time transmission, a fast chaotic image encryption algorithm using a novel divide and conquer diffusion strategy is proposed in this paper. Firstly, a fast pseudo-random sequence generator is constructed using 2D hyperchaotic systems. And by performing bitwise XOR operation between two integer sequences obtained from different systems, the final key stream used for diffusion will have better randomness and unpredictability. Secondly, to achieve divide and conquer diffusion, the plain image is divided into three parts, then a parallel CBC-based diffusion method can simultaneously act on upper and bottom parts (or left and right parts), which achieves a time complexity of $\boldsymbol {O}$ ( $W+H$ ). Moreover, within the diffusion process, the mechanism of information interaction on the central row or column further enhances avalanche effect to resist differential attack. Thirdly, the session key of the proposed algorithm is a mixture of the plain image’s hash value and a true random sequence, which not only improve plaintext sensitivity but also realize one-time pad. Finally, experimental results indicate the superiority of our algorithm to resist statistical, chosen-plaintext, entropy, and other common attacks. Furthermore, by comparison with previous works, our algorithm preforms much faster execution speed with only average of 0.08s to encrypt images of size $512\times 512$ , while it provide the same or even higher level of security. Therefore, the proposed algorithm can meet security and efficiency requirements of real-time communications of image data.
12 March 2022 in Journal of Information Security and Applications
With the application of new technologies in vehicles such as wireless communications, microelectronics, Artificial Intelligence (AI) and Autonomous Vehicles (AV), there are lots of embedded hardware and software in vehicles, and so cryptographic technologies such as block ciphers are also applied to the vehicle system to protect its security and privacy. In order to better find potential security vulnerabilities and improve security of the vehicle system which applied SIMON block cipher, in this paper, a novel Differential fault attack (DFA) method is presented. By inducing a random bit fault in $L^{T-4}$ (the fourth round from the last), we first demonstrate the process of identifying the fault-induced position. Then, on the basis of the determined position, we describe how to retrieve three rounds of keys by inducing only one round of faults. For retrieving the SIMON family, compared with previous DFAs under the random bit fault model, the lowest numbers of fault-induced and fault locations (or rounds) are required by our attack. In particular, when the key words m is 2 or 3, the proposed attack only needs to induce one round of faults. Finally, we carry out simulation verification and a comparison to illustrate the correctness and effectiveness of the proposed attack.
Differential fault attack (DFA) is a distinctive methodology for acquiring the key to block ciphers, which comprises two distinct strategies: DFA on the state and DFA on the key schedule. Given the widespread adoption of the Advanced Encryption Standard (AES), it has emerged as a prominent target for DFA. This paper presents an efficient DFA on the AES, utilizing a two−byte fault model that induces faults at the state with discontiguous rows. The experiment demonstrates that, based on the proposed fault model, the key for AES–128, AES–192, and AES–256 can be successfully recovered by exploiting two, two, and four faults, respectively, without the need for exhaustive research. Notably, in the case of AES–256, when considering exhaustive research, two (or three) faults are needed with 232 (or 216) exhaustive searches. In comparison to the currently available DFA on the AES state, the proposed attack method shows a higher efficiency due to the reduced induced faults.