Showing papers on "Bit plane published in 2020"

A novel image encryption algorithm based on bit-plane matrix rotation and hyper chaotic systems

Abstract: In this paper, we propose a new image encryption algorithm based on bit-plane matrix rotation and two hyper chaotic systems. The algorithm first decomposes the plain-image into eight bit planes and constructs a three-dimensional (3D) matrix. Then the sub-matrix of the 3D bit-plane matrix is rotated in different directions controlled by PRNS generated by a hyper-chaotic system. Finally, the pixel values of the intermediate image are modified by using another key stream. Furthermore, the initial values of diffusion and parameters related with generating chaotic sequences are produced by the MD5 hash function of the plain-image, which enhances the correlation between the encryption process and the plain-image. Simulation experiments are presented to analyze the image encryption scheme in terms of key space, histogram, information entropy, key sensitivity and adjacent pixels correlation index. Theoretical analysis and experimental results demonstrate that the proposed algorithm has excellent performance and suffcient security level.

Efficient image encryption scheme based on generalized logistic map for real time image processing

Abstract: In this era of the information age with digitalization, the transmission of sensitive real-time image information over insecure channels is highly-likely to be accessed or even attacked by an adversary. To prevent such unauthorized access, cryptography is being used to convert sensitive information in real-time images into unintelligible data. Most of the time, schemes are proposed with a high level of security. However, the challenge always remains the slower speeds due to their high complexity which makes them unusable in the applications of real-time images. In this paper, an efficient image encryption algorithm has been developed and tested for real-time images. The proposed scheme makes use of encryption with an efficient permutation technique based on a modular logistic map to bring down the size of the chaotic value vector, required to permute real-time image. We show that an efficient permutation is obtained using only $$\sqrt{N}$$ chaotic numbers for a square image with 3N pixels (N Pixels in each color bit plane). The algorithm makes use of a 192-bit key; divided into smaller blocks and each block selected chaotically to diffuse the pixel using multiple XOR operations. The experimental analysis reveals that the proposed algorithm is immune to various statistical and differential attacks such as entropy, histogram analysis, spectral characteristic analysis, etc. A comparison of the proposed scheme with some state-of-the-art techniques show that it performs better, and as such, can be utilized for efficient real-time image encryption.

Reversible Data Hiding in Encrypted Images Based on Pixel Prediction and Bit-plane Compression

Abstract: Reversible data hiding in encrypted images (RDHEI) receives growing attention because it protects the content of the original image while the embedded data can be accurately extracted and the original image can be reconstructed losslessly. To make full use of the correlation of the adjacent pixels, this paper proposes an RDHEI scheme based on pixel prediction and bit-plane compression. Firstly, the original image are divided into the blocks of equal size and the prediction error of the original image is calculated. Then, the 8 bit-planes of prediction error are executed rearrangement and bit-stream compression, respectively. Finally, the image after vacating room is encrypted by a stream cipher and the additional data is embedded in the vacated room by multi-LSBs (Least Significant Bits) substitution. Experimental results show that the embedding capacity of the proposed method outperforms the state-of-the-art methods.

Local bit-plane decoded convolutional neural network features for biomedical image retrieval

Abstract: Biomedical image retrieval is a challenging problem due to the varying contrast and size of structures in the images. The approaches for biomedical image retrieval generally rely on the feature descriptors to characterize the images. The feature descriptor of query image is compared with the descriptors of images from the database, to find the best matches. Several hand-crafted feature descriptors have been proposed so far for biomedical image retrieval by exploiting the local relationship of neighboring image pixels. It is observed in the literature that the local bit-plane decoded features are well suited for this retrieval task. Moreover, in recent past, it is also observed that the convolutional neural network-based features such as AlexNet, Vgg16, GoogleNet and ResNet perform well in many computer vision-related tasks. Motivated by the success of the deep learning-based approaches, this paper proposes a local bit-plane decoding-based AlexNet descriptor (LBpDAD) for biomedical image retrieval. The proposed LBpDAD is computed by max-fusing the ReLU operated feature maps of pre-trained AlexNet at a particular layer, obtained from the original and local bit-plane decoded images. The proposed approach is also compared with Vgg16, GoogleNet and ResNet models. The experiments on the proposed method over three benchmark biomedical databases of different modalities such as MRI, CT and microscopic show the efficacy of the proposed descriptor.

Reversible data hiding in encrypted images based on pixel prediction and multi-MSB planes rearrangement

Abstract: Great concern has arisen in the field of reversible data hiding in encrypted images (RDHEI) due to the development of cloud storage and privacy protection. RDHEI is an effective technology that can embed additional data after image encryption, extract additional data error-free and reconstruct original images losslessly. In this paper, a high-capacity and fully reversible RDHEI method is proposed, which is based on pixel prediction and multi-MSB (most significant bit) planes rearrangement. First, the median edge detector (MED) predictor is used to calculate the predicted value. Next, unlike previous methods, in our proposed method, signs of prediction errors (PEs) are represented by one bit plane and absolute values of PEs are represented by other bit planes. Then, we divide bit planes into uniform blocks and non-uniform blocks, and rearrange these blocks. Finally, according to different pixel prediction schemes, different numbers of additional data are embedded adaptively. The experimental results prove that our method has higher embedding capacity compared with state-of-the-art RDHEI methods.

Searchable Encrypted Image Retrieval Based on Multi-Feature Adaptive Late-Fusion

Abstract: Recently, searchable encrypted image retrieval in a cloud environment has been widely studied However, the inappropriate encryption mechanism and single feature description make it hard to achieve the expected effects Therefore, a major challenge of encrypted image retrieval is how to extract and fuse multiple efficient features to improve performance Towards this end, this paper proposes a searchable encrypted image retrieval based on multi-feature adaptive late-fusion in a cloud environment Firstly, the image encryption is completed by designing the encryption function in an RGB color channel, bit plane and pixel position of the image Secondly, the encrypted images are uploaded to the cloud server and the convolutional neural network (CNN) is fine-tuned to build a semantic feature extractor Then, low-level features and semantic features are extracted Finally, the similarity score curves of each feature are calculated, and adaptive late-fusion is performed by the area under the curve A large number of experiments on public dateset are used to validate the effectiveness of our method

Saliency-based bit plane detection for network applications

Abstract: Transmitting image data without losing significant information is challenging for any network application especially when large color images are transmitted through TCP communication protocol. This is due to network limitations such as buffer overflow, underflow and network traffic flow etc. This paper presents a new method for image size reduction such that the network can transmit data without much loss of information, and hence, quality. The proposed method obtains bit planes for the color input images, which results in eight binary planes. Unlike the existing bit plane based image size reduction methods, which assume that the most significant plane or some other planes contain useful information, the proposed method finds the plane that contains dominant information automatically. For each plane, the proposed method explores the saliency that finds dominant information based on Markov Chain Process and similarity estimation between neighbor pixels. To reduce computational burden, we use Canny edge maps of the saliency of the planes for feature extraction. We propose to explore ring-growing concept for the edge maps to study the spatial distribution of saliency, locally. The proposed method detects the plane based on statistics of saliency distribution. To validate the step of plane detection, we estimate transmission error caused during data transmission through TCP communication protocol for the images at sending and receiving ends. Experimental results on plane detection show that the proposed method is better than the existing methods in terms of detection rate. Our experiments on image transmission through TCP communication protocol show that the proposed method outperforms the existing methods in terms of error estimation and quality analysis. Furthermore, experiments are conducted to analyze packet loss in terms of number of duplicate acknowledgements and retransmission during packets transmission for the color, edge and plane to show that transmitting plane images improves network performance in terms of less number of duplicate acknowledgement, retransmission and time taken in seconds.

A superlative image encryption technique based on bit plane using key-based electronic code book

Abstract: To contend with the growing concern of information security, firms are espousing cryptography for protection. To comprehend this need, the paper proposes a block cipher that ensures confidentiality and secrecy for secure data communication network. A high-quality cryptographic process ensures high entropy, high key sensitivity, ability to resist known plaintext and chosen-plaintext attack, high speed of execution, high key space, high randomness and resistance towards differential attack. In projected work, encryption mechanism is comprises of multiple processes and each process is devised key dependent utilizing diverse keys to ensure high key sensitivity and vast resistance power towards differential attack. The keys are generated using contemporary encryption process- the quantum chaotic map due to its high randomness and non periodicity, which includes confusion and diffusion processes. For the confusion process, Electronic Code Book (ECB), Initial Permutation (IP), Bit plane scrambling, and Inter bit plane scrambling are employed. The ECB and IP, being matching processes are chosen for high speed of execution. Bit level permutation unlike byte level is applied to reinforce randomness, in conjunction with variable number of rounds as per the security key. For diffusion process, the folding technique is used along eight directions, exploiting different keys. To reveal the efficacy of the proposed methodology, it is compared with the existing renowned methods on the basis of Cryptanalysis, Perceptibility analysis, Statistical analysis, etc. It is asserted that results such as differential attack, cryptanalysis, and key space analysis are reasonably enhanced than others while entropy and speed of execution are at equivalence with other techniques.

A New Bit Plane Specific Longest Repeating Pattern Test for Statistical Analysis of Bit Sequences

Abstract: Most of the existing randomness tests treat whole bit sequence as a stream of bits in one-dimensional space for analysis. In this paper, a new bit plane specific statistical test for randomness analysis of bit sequences in two-dimensional space is presented. Primary focus of this test is to analyze the length of longest repeating pattern (LRP) in a binary random sequence. A random sequence is converted into matrix of L × L bytes. Then analysis of LRP is done for each row and column of each bit plane of the matrix. Experimentations are carried out on cipher data of four block ciphers, image data, and random alphabetic data of various sizes. It is observed that cipher data exhibits good randomness but image and random alphabetic data exhibits non-randomness in bit planes. The proposed test has the capability to capture the patterns by performing two-dimensional analysis in bit planes which may not be possible to capture by existing NIST statistical tests. It has been shown that sequences accepted as random by NIST suite are rejected by proposed LRP test.

ABMC-DH: An Adaptive Bit-Plane Data Hiding Method Based on Matrix Coding

Abstract: This paper proposes a novel adaptive image steganography method combining matrix coding, which can achieve better visual quality according to the given embedding rate compared with other existing schemes. The embedding processes are divided into two phases: preparation and implementation. In the preparation phase, all 128 combinations of a 7-bit binary number are classified into eight groups by a parity check matrix based on matrix coding. Thus, each group corresponds to an octal number, and contains 16 candidates. Correspondingly, all pixels of a cover image are classified into three categories by a predefined embedding rate level: low, middle, and high. In the implementation phase, each pixel in the high level can be embedded with 3 bits by substituting its 7 LSBs (Least Significant Bit) using a close candidate from the group determined by the to-be-embedded 3 bits. For the middle level, pixel pairs are used to hide 3-bit data by substituting its (3, 4) LSBs combination using a close candidate form the group determined by the to-be-embedded 3 bits. For pixels in the low level, a triple pixel combination is used to embed 3-bit data by substituting its (2, 2, 3) LSBs combination using a close candidate form the group determined by the to-be-embedded 3 bits. The experimental results indicate that the proposed scheme is optimal compared to other existing matrix-coding data hiding schemes in terms of visual quality, while providing an identical embedding capacity.

Research on high frame rate scene projection method based on digital micromirror device (DMD)

Abstract: The digital micromirror device DMD is widely used in visible light projection, special-purpose spatial light modulation, and infrared scene simulation, due to its high resolution, uniformity and energy concentration. In some applications that require high frame rate scene image display, it is necessary to ensure that the DMD displays high gray level images at a high frame rate. However, the display frame rate of the pulse width modulation (PWM) method is limited by the minimum time required for DMD loading data, unable to achieve high frame rate display. Although the DMD binary display mode can meet the requirements of high frame rate display, it cannot meet the requirements of high gray level. The paper proposes an image display method that uses light source and DMD to synchronize modulation. Decompose the high-order gray image into bit planes according to the gray threshold, and the DMD displays each bit plane in binary mode under the trigger of an sync pulse. The intensity of the illumination laser source is modulated by an acousto-optic modulator to match the bit plane and the radiated laser power. This method makes use of the function of high frame rate display in DMD binary mode cooperate with light intensity modulation of illumination laser source, and realizes high frame rate and high dynamic range image display. With this method, the maximum frame rate of 8-bit gray level image with 1920 ×1080 resolution can reach 2KHz. The experimental system has realized 200Hz frame rate display of 8-bit gray level image with 1920 ×1080 resolution.

Data hiding in virtual bit-plane using efficient Lucas number sequences

Abstract: This paper introduces a steganography technique using the concept of virtual bit-plane. Purposely another number system has been used in this technique instead of binary to hide the target data. After that embedding of target data is done on different bit-planes. Here for conversion, Lucas Number system is used. The Lucas sequence is almost similar to Fibonacci, instead 1 and 2 it starts with 2 and 1. It helps to increase robustness by embedding data at the second bit-plane with a slight change of ±1. Here for embedding target data, Blue and Green channels of RGB color image are used. Red channel is used as indicator for proper extraction of target data at the receiver side. The indicator is used to avoid lexicographically higher order to consider for number representation. It may seem that the use of two channels for embedding reduces the capacity. But it does not really happen. The skip of pixel to follow Zekendrof’s rule for handling redundant representation by other existing methods, make the proposed one more capacitive than other. In order to establish its efficiency over the state-of-art-works the proposed method is analyzed by using different parameters and compared with relevant techniques. It has been found from the tested result that the proposed one is better. The stego quality of the method is also maintained along with the robustness and capacity. The PSNR of the proposed method is within the acceptable range, even in the highest embedding capacity.

Data compression method and device, data decompression method and device and processing method and device based on data compression and decompression

Abstract: The invention provides a data compression method and device, a data decompression method and device and a processing method and device based on data compression and decompression, and the data compression method comprises the steps: receiving to-be-compressed data which is sparse data outputted by any layer of a neural network model; and performing compression processing on the to-be-compressed data based on a sparse compression algorithm and a bit plane compression algorithm to obtain compressed data By utilizing the method, a relatively high compression ratio can be realized, so that the data transmission bandwidth and the storage space of an external memory can be saved, the memory access efficiency is improved, and the chip computing power is improved

High Data Rate Audio Steganography

Abstract: Steganography is an art to hide information without affecting perceptual transparency of digital media files. Media files include audio, video, image, text, software and so on. Hiding information is required for secure transmission of data. In general, media files are known as cover file which wrap the data to be transmitted. The media file must be imperceptible before and after embedding of data. But when we embed data, we do some modification at different bit plane that introduces error or noisy media. Motivation of this work is based on the property of human auditory system (HAS) which is discernible towards any kind of modification or distortion in audio file. Keeping in mind this HAS property, in this work, we concentrate on hiding a large amount of data in a small size cover file considering reduced noise and minimized bit error rate.

A simple encoder scheme for distributed residual video coding

Abstract: Rate-Distortion (RD) performance of Distributed Video Coding (DVC) is considerably less than that of conventional predictive video coding. In order to reduce the performance gap, many methods and techniques have been proposed to improve the coding efficiency of DVC with increased system complexity, especially techniques employed at the encoder such as encoder mode decisions, optimal quantization, hash methods etc., no doubt increase the complexity of the encoder. However, low complexity encoder is a widely desired feature of DVC. In order to improve the coding efficiency while maintaining low complexity encoder, this paper focuses on Distributed Residual Video Coding (DRVC) architecture and proposes a simple encoder scheme. The main contributions of this paper are as follows: 1) propose a bit plane block based method combined with bit plane re-arrangement to improve the dependency between source and Side Information (SI), and meanwhile, to reduce the amount of data to be channel encoded 2) present a simple iterative dead-zone quantizer with 3 levels in order to adjust quantization from coarse to fine. The simulation results show that the proposed scheme outperforms DISCOVER scheme for low to medium motion video sequences in terms of RD performance, and maintains a low complexity encoder at the same time.

Chaos-based novel parallel image encryption method

Abstract: The invention discloses a chaos-based novel parallel image encryption method The method comprises the following steps of 1) controlling an initial condition of a chaos system by using a proposed adaptive adjustment parameter; 2) generating a corresponding chaotic sequence, and quantizing the chaotic sequence to construct an encryption matrix; and 3) obfuscating the plaintext image on the bit plane, then combining a dynamic index diffusion method with the encryption matrix to be applied to a diffusion process, and diffusing the obfuscated image to obtain a final encrypted image According to the method, parallel acceleration is realized based on a graphic processing unit (GPU), the encryption efficiency can be improved, various attacks based on cryptographic analysis can be effectively resisted, and the method has high encryption speed and good security performance

Reversible Data Hiding in Encrypted Images Based on Bit plane Compression of Prediction Error.

Abstract: As a technology that can prevent the information of original image and additional information from being disclosed, the reversible data hiding in encrypted images (RDHEI) has been widely concerned by researchers. How to further improve the performance of RDHEI methods has become a focus of research. To this end, this work proposes a high-capacity RDHEI method based on bit plane compression of prediction error. Firstly, to reserve the room for embedding information, the image owner rearranges and compresses the bit plane of prediction error. Next, the image after reserving room is encrypted with a serect key. Finally, the information hiding device embeds the additional information into the reserved room. This paper makes full use of the correlation between adjacent pixels. Experimental results show that this method can realize the real reversibility and provide higher embedding capacity than state-of-the-art works.

Image Encryption Using Genetic Algorithm and Bit-Slice Rotation

Abstract: Cryptography is a powerful means of delivering information in a secure manner. Over the years, many image encryption algorithms have been proposed based on the chaotic system to protect the digital image against cryptography attacks. In chaotic encryption, it jumbles the image to vary the framework of the image. This makes it difficult for the attacker to retrieve the original image. This paper introduces an efficient image encryption algorithm incorporating the genetic algorithm, bit plane slicing and bit plane rotation of the digital image. The digital image is sliced into eight planes and each plane is well rotated to give a fully encrypted image after the application of the Genetic Algorithm on each pixel of the image. This makes it less prone to attacks. For decryption, we perform the operations in the reverse order. The performance of this algorithm is measured using various similarity measures like Structural Similarity Index Measure (SSIM). The results exhibit that the proposed scheme provides a stronger level of encryption and an enhanced security level.

Технологія кодування трансформант в системах управління бітовою швидкістю відеопотоку

Abstract: The conceptual basements of constructing an effective encoding method within the bit rate control module of video traffic in the video data processing system at the source level are considered. The essence of using the proposed method in the course of the video stream bit rate controlling disclosed, namely, the principles of constructing the fragment of the frame code representation and approaches for determining the structural units of the individual video frame within which the control is performed. The method focuses on processing the bit representation of the DCT transformants, and at this processing stage transformant was considered as a structural component of the video stream frame at which the encoding is performed. At the same time, to ensure the video traffic bit rate controlling flexibility, decomposition is performed with respect to each of the transformants to the level of the plurality of bit planes. It is argued that the proposed approach is potentially capable to reducing the video stream bit rate in the worst conditions, that is, when component coding is performed. In addition, this principle of video stream fragmen code representation forming allows to control the level of error that can be made in the bit rate control process. However, in conditions where the bit representation of the transformant is encoded, the method is able to provide higher compression rates as a result of the fact that the values of the detection probability of binary series lengths and the values of detected lengths within the bit plane will be greater than in the case of component coding. This is explained by the structural features of the distribution of binary elements within each of the bit planes, which together form the transformer DCT. In particular, high-frequency transformer regions are most often formed by chains of zero elements. The solutions proposed in the development of the encoding method are able to provide sufficient flexibility to control the bit rate of the video stream, as well as the ability to quickly change the bit rate in a wide range of values.

Image encoding method and decoding method, and device and system to which said methods are applicable

Abstract: The present application provides an image encoding method and decoding method, and a device and a system to which said methods are applicable. Said encoding method comprises: dividing acquired image data into a plurality of bit plane matrix data according to binary data bits; performing serialization processing on bit plane matrix data of at least a part of bit planes on the basis of a preset serialization period to obtain bit plane serialized data, the serialization period being a period set by serializing a preset m*n matrix according to neighboring data; and encoding the obtained bit plane serialized data, and generating encoded image data of the image data. The present application uses a serialization period to implement serialization processing on bit plane matrix data, and is beneficial in improving the cohesiveness of an original image, particularly the cohesiveness of high definition images of 4K or more.

Method for efficient encoding and decoding of a digital image

Abstract: FIELD: encoding and decoding devicesSUBSTANCE: invention relates to means for encoding and decoding digital images A digital image is read from the array of external memory in the form of a two-dimensional block Cof size N= N× N, where Nis the number of pixels in a row/column, to obtain a one-dimensional block which is stored in a register encoder memory Simultaneously, the coder calculates the number of the upper bit plane MSB of the two-dimensional block C, analyzes the signs of its significant pixels S, from the one-dimensional block, the bit planes BPare identified, where n is from 0 to MSB, which are further sequentially and independently from each other treated For each bit plane, forming L-level hierarchical binary quad tree m, where L = logN, wherein zero level is formed by copying all bits of bit plane BP, and subsequent levels are filled on the basis of previous by estimation of significance of their bits according to formulawherefurther compactness of map of significance Cfor obtained L-level hierarchical binary quadrupleEFFECT: technical result is higher efficiency of digital image processing2 cl, 3 dwg

Method and device for fast and efficient compression and decompression of images

Abstract: FIELD: data processing.SUBSTANCE: present invention relates to compression and decompression of data, in particular, for large images. Said result is achieved by encoding with transformation of image data to form a visual representation using transformation coefficients, serializing visual representation bit planes using transform coefficients and optimum prefix bit encoding in a bit plane for each bit plane, having a common local context, encoding a series of zero bit sequences in the bit plane with lengths and storing coefficients, received after optimal prefix encoding and encoding the zeros series of symbols, starting from the sign and further bits in order of significance, starting from the most significant bit to further to the least significant bit in the lookup table in the header section.EFFECT: technical result is provision of fast and efficient compression providing dynamic trade-off between use of bandwidth and image quality.14 cl, 3 dwg

Ultra-high-definition video compression encoding and decoding method with ultra-low time delay

Abstract: The invention belongs to the technical field of wireless multimedia communication, and discloses an ultra-high-definition video compression encoding and decoding method with ultra-low time delay, which comprises the following steps that: an encoding end performs input preprocessing on an input video image to remove a direct-current component; asymmetric wavelet decomposition is carried out on thepreprocessed image to obtain sub-bands with different resolutions, and proportional extraction is carried out on wavelet coefficients after transformation in a frequency domain; the processed waveletcoefficients are reasonably divided into regions, and code rate control is performed on the regions to obtain corresponding quantization parameters and entropy coding modes; entropy coding and code stream organization are carried out on a quantized code block coefficient; at a decoding end, firstly, header analysis is carried out on an obtained code stream to obtain decoding parameters; entropy decoding is carried out to obtain wavelet coefficient values and bit plane counting positions, and interpolation recovery is carried out after inverse quantization; and inverse wavelet transform is performed to obtain original image data. According to the method, the complexity of compression and decompression algorithms is effectively reduced, and meanwhile, efficient compression and decompressionand low-delay output of an ultra-high-definition video are realized.

Bit plane decoding method and apparatus

Abstract: The present application provides a bit plane decoding method. A processing mode of "prediction + reconstruction" is adopted in the bit plane decoding. The vast majority of positions that do not need to be decoded can be skipped in the reconstruction process by means of one or more predictions. Compared with the existing "scanning reconstruction" mode, the present application reduces a lot of idle time in the reconstruction process and increases the speed of bit plane decoding. In addition, the area consumption caused by the degree of parallelism of a bit plane decoder (BPD) can also be reduced, and the decoding performance can be improved as a whole.

Encoding method, encoder, and encoding system

Abstract: Provided are an encoding method, an encoder, and an encoding system. The encoder comprises: acquiring zero bit plane (ZBP) information or inclusion information of each code block to be encoded in a set of code blocks to be encoded; and constructing an identifier tree according to the ZBP information or inclusion information of each code block, and encoding the identifier tree in parallel.

Complex Method of Video Stream Intensity Control

Abstract: Consider the construction of a video stream bit rate controlling method, based on the non-equilibrium positional coding of a transformants bit representation. Developed the concept of combined use of technologies, which allows to adapt the video data intensity to the channel bandwidth in the conditions of its dynamic change.

Bit plane encoding of data arrays

Abstract: Disclosed herein is a method of encoding an array of data elements comprising transforming the array from the spatial to the frequency domain, representing the frequency domain coefficients as a plurality of bit plane arrays, and encoding the set of frequency domain coefficients as a data packet having a fixed size by encoding the bit plane arrays in a bit plane sequence working from the bit plane array representing the most significant bit downwards until the data packet is full. Each bit plane array is encoded by recursively subdividing the bit plane array into respective sections and subsections down to the individual coefficients and including in the data packet, so long as there is available space, data indicating the locations of any (sub)sections in that bit plane array that for the first time in the bit plane sequence contain one or more coefficient(s) having a non-zero bit value.

Defending Against Adversarial Attacks in Deep Learning with Robust Auxiliary Classifiers Utilizing Bit Plane Slicing

Abstract: Deep Neural Networks (DNNs) have been widely used in variety of fields with great success. However, recent researches indicate that DNNs are susceptible to adversarial attacks, which can easily fool the well-trained DNNs without being detected by human eyes. In this paper, we propose to combine the target DNN model with robust bit plane classifiers to defend against adversarial attacks. It comes from our finding that successful attacks generate imperceptible perturbations, which mainly affects the low-order bits of pixel value in clean images. Hence, using bit planes instead of traditional RGB channels for convolution can effectively reduce channel modification rate. We conduct experiments on dataset CIFAR-10 and GTSRB. The results show that our defense method can effectively increase the model accuracy on average from 8.72% to 85.99% under attacks on CIFAR-10 without sacrificina accuracy of clean images.