LSB嵌入方法的载体图像区域选择

针对在密写术中的LSB（最低有效位）嵌入方法会对载体图像的平滑区域造成破坏,导致载体图像的不可感知性下降和不可检测性降低的问题,提出了一种区域选择方法,避免载体图像的脆弱区域在LSB嵌入操作中被破坏。通过实验证明了,该方法可以有效改善视觉不可感知性,提高对LSB嵌入操作的不可检测性。     

联合多重对抗与通道注意力的高安全性图像隐写

目的 现有基于对抗图像的隐写算法大多只能针对一种隐写分析器设计对抗图像,且无法抵御隐写分析残差网络(steganalysis residual network,SRNet)、Zhu-Net等最新基于卷积神经网络隐写分析器的检测。针对这一现状,提出了一种联合多重对抗与通道注意力的高安全性图像隐写方法。方法 采用基于U-Net结构的生成对抗网络生成对抗样本图像,利用对抗网络的自学习特性实现多重对抗隐写网络参数迭代优化,并通过针对多种隐写分析算法的对抗训练,生成更适合内容隐写的载体图像。同时,通过在生成器中添加多个轻量级通道注意力模块,自适应调整对抗噪声在原始图像中的分布,提高生成对抗图像的抗隐写分析能力。其次,设计基于多重判别损失和均方误差损失相结合的动态加权组合方案,进一步增强对抗图像质量,并保障网络快速稳定收敛。结果 实验在BOSS Base 1.01数据集上与当前主流的4种方法进行比较,在使用原始隐写图像训练后,相比于基于U-Net结构的生成式多重对抗隐写算法等其他4种方法,使得当前性能优异的5种隐写分析器平均判别准确率降低了1.6%;在使用对抗图像和增强隐写图像再训练后,相比其他4种方法,仍使得当前性能优异的5种隐写分析器平均判别准确率降低了6.8%。同时也对对抗图像质量进行分析,基于测试集生成的2 000幅对抗图像的平均峰值信噪比(peak signal-tonoise ratio,PSNR)可达到39.925 1 dB,实验结果表明本文提出的隐写网络极大提升了隐写算法的安全性。结论 本文方法在隐写算法安全性领域取得了较优秀的性能,且生成的对抗图像具有很高的视觉质量。     

抗统计分析的BPCS隐写算法

位平面复杂度分割（BPCS）隐写在复杂度较高的位面小块嵌入秘密信息,有较好的隐蔽性,但会引起图像位面小块复杂度的统计特性的变化。通过深入分析图像位平面上相邻像素之间的相关性,本文提出了一种改进的BPCS隐写法。它更好地利用人的视觉特性,结合图像的局部特征来嵌入秘密信息。实验结果表明,改进后的算法能够抵御针对原BPCS隐写的分析方法,有较大的嵌入量,并对原图像所作的改变具有更好的不可见性。     

基于小波包分解和支持向量机的图像隐秘检测算法

提出了一种新的小波包分解算法,该算法只对图像的低频部分进行分解,从而大大减少了特征向量的维数,提高了运行效率。在模式分类上采用了支持向量机的算法,实验证明,该算法与Fisher线性判别算法相比,其检测率明显提高。     

一种新的基于H.264/AVC的可逆鲁棒隐写方法

提出了一种新的基于H。264/AVC的可逆鲁棒视频隐写方法。该方法首先利用（t，n）门限秘密共享技术将需要嵌入的秘密信息用多项式分发获得子秘密，再利用BCH码对子秘密进行编码，然后通过H。264的预测模式选择嵌入位置，最后将编码后的子秘密嵌入到视频的各个帧中。提取时，利用视频中t个编码后的子秘密恢复秘密信息。实验结果标明，该方法在提取秘密信息时能完全恢复原视频，具有较强的比特错、帧错鲁棒性，在随机丢帧率小于等于15%的情况下能完全恢复秘密信息，能效控制帧内失真漂移，视觉效果良好。     

A steganographic method based on optimized audio embedding technique for secure data communication in the internet of things

The rapid growth of the internet and the internet of things (IoT) refers to the next phase of information revolution whose context involves billions of smart devices and sensors interconnected to facilitate speedy information and data exchange under soft real-time constraints. Digital information revolution has caused significant changes in the data communication. This data communication may require private, secure, and sometimes malicious communication. Competent secrecy can be accomplished by applying novel and inventive audio steganography. This article focuses on the secret message followed by shuffled embedded bit substitution in original audio stream by adopting optimized audio embedding technique (OAET) from the technological observation. To hide the information in the deeper layer of the audio stream, this method uses a new elevated bit range least significant bit (LSB) audio steganography technique that decreases distortion and improves the robustness of the embedded audio stream. The proposed technique proves that the perceptual quality of audio steganography is better than the previous standard LSB technique. Experiment results proved that the cladding of the OAET provides high-level security to the universal cyber data. The interpretation of results shows that embedding data in audio enhances the level of security when used as IoT smart speakers, where the attackers could not distinguish between the original audio and the embedded audio streams.     

Advancing DNA Steganography with Incorporation of Randomness

DNA has become a promising candidate as a future data storage medium; this makes DNA steganography indispensable in DNA data security. PCR primers are conventional secret keys in DNA steganography. Brute force testing of different primers will be extremely time consuming, and practically unaffordable when high-throughput sequencing is used. However, the encrypted information can be sequenced and read once the primers are intercepted. A new steganography approach is needed to make the DNA-encoded information safer, if not unhackable. Mixing information-carrying DNA with a partially degenerated DNA library containing single or multiple restriction sites, we have built an additional protective layer that can be removed by desired restriction enzymes as secondary secret keys. As PCR is inevitable for reading DNA-encrypted information, heating will cause reshuffling and generate endonuclease-resistant mismatched duplexes, especially for DNA with high sequence diversity. Consequently, with the incorporation of randomness, DNA steganography possesses both quantum key distribution (QKD)-like function for detecting PCR by an interceptor and a self-destructive property. It is noteworthy that the background noise generated through the protective layer is independent from any sequencing technology including Sanger and high-throughput sequencing. With a DNA ink incorporating the steganography, we have shown that the authenticity of a piece of writing can be confirmed only by authorized persons with knowledge of all embedded keys.     

Memetic Optimization with Cryptographic Encryption for Secure Medical Data Transmission in IoT-Based Distributed Systems

In the healthcare system, the Internet of Things (IoT) based distributed systems play a vital role in transferring the medical-related documents and information among the organizations to reduce the replication in medical tests. This datum is sensitive, and hence security is a must in transforming the sensational contents. In this paper, an Evolutionary Algorithm, namely the Memetic Algorithm is used for encrypting the text messages. The encrypted information is then inserted into the medical images using Discrete Wavelet Transform 1 level and 2 levels. The reverse method of the Memetic Algorithm is implemented when extracting a hidden message from the encoded letter. To show its precision, equivalent to five RGB images and five Grayscale images are used to test the proposed algorithm. The results of the proposed algorithm were analyzed using statistical methods, and the proposed algorithm showed the importance of data transfer in healthcare systems in a stable environment. In the future, to embed the privacy-preserving of medical data, it can be extended with blockchain technology.     

基于辨识性统计特征的PQ隐密图像识别算法

提出一种基于辨识性统计特征的PQ（perturbed quantization）隐密图像识别算法。该算法根据经典PQ隐写对图像数据的更改方式,提取可有效区分该类隐密图像与其他类隐密图像的辨识性统计特征,并运用SVM（support vector machines）分类器进行分类识别。实验结果表明：本算法能够可靠地将PQ隐密图像从5类典型JPEG隐写PQ、F5、nsF5、MB1和MOD的隐密图像中识别出来;即使F5、nsF5、MB1和MOD的隐密图像不参与分类器的训练,本算法仍能有效识别PQ隐密图像。     

多媒体隐写研究进展

大数据分析可以跳过数据内容而仅从数据背景挖掘情报,传统的加密通信已经难以满足安全通信的需求。隐写技术是将秘密消息嵌入各种载体(如数字图像、音频、视频或文本)中实现隐蔽通信的技术,是应对大数据情报获取的有效手段,是密码技术的必要补充。人工智能,尤其是深度学习,在计算机视觉、语音和自然语言处理等领域的巨大成功,给隐写术带来了新机遇,提出了新挑战,促使基于图像、音/视频和文本的隐写术出现了一系列新思想、新方法。本文介绍隐写术的概念、分类、主要作用和研究意义,概述隐写术的发展历史、研究近况和应用场景。注意到各类载体上的隐写术虽然有差别,但是其核心追求有共通之处,可以提炼成通用的隐写编码问题。所以本文首先介绍隐写编码的基本思想与关键技术,然后针对最重要和流行的载体、图像、视频、音频和文本,分别介绍隐写术的进展。总体而言,本文从隐写编码、图像隐写、视频隐写、音频隐写和文本隐写5个方面概述隐写术的国际/国内发展现状,总结差异,对比优势和劣势,并分析发展趋势。