BLIND AUDIO WATERMARKING BY HYBRID METHOD WITH HIGH IMPERCEPTIBILITY, ROBUSTNESS AND CAPACITY

• Main Author: Budiman, Gelar
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55528
• Institution: Institut Teknologi Bandung

Insertion or hiding of data / information in a content is one technique to improve data security. In the current era this is where data transactions are spread so easily in the internet world the chance of illegally distributing content is higher. In this research content which will be studied focused on audio content. Audio content in the form of the work the copyright of a producer or creator is often misused with distribution illegal content in order to gain personal profit so necessary a watermark hiding technique that contains copyright from the creator or audio producers are called audio watermarking. Watermark hiding techniques in an audio must meet the criteria watermark concealment: imperceptible, resistant to audio damage caused many ways of content destruction, and relatively high watermark data capacity. Many methods have been used for hiding watermark data on audio. All data hiding methods are divided into 4 types of methods include: time domain, frequency domain, other domains, and hybrid methods. Hybrid method is a combination of watermark data insertion methods on audio, for example by combining one time domain method with one frequency domain method. In this dissertation there are several insertion techniques developed for reach the watermarking criteria. Insertion techniques developed in This dissertation includes Multicarrier (MC), Spread Spectrum (SS) with insertion of single and plural bits, Quantization Index Modulation (QIM), and modification of magnets Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT) with Fibonacci. Before doing the insertion with several techniques the processing is carried out on the audio host so that it is obtained strong sample that represents one audio segment. Watermark embedding using the SS method is done on several research, including the insertion of single bits in the time domain of audio with SS modulation modification, insertion of single bits in the frequency domain of the subband certain audio with Lifting Wavelet Transform (LWT) and FFT and insertion bit plural on certain audio subband with Stationary Wavelet decomposition Transform (SWT). Bit insertion technique in one of the studies combined with the process of data compression using the Compressive Sampling technique (CS) to obtain a code sample from compressed audio and contains an inserted watermark. The next research is concealment watermark with magnetic modification from FFT and DCT ivresults utilizing a series Fibonacci and compare insertion performance with FFT and DCT. Furthermore, research by inserting data using the MC method was conducted in the time domain of audio by modifying the MC in the bit inserted, filtering with psychoacoustic filters, then added in segment-based audio after adjusting with the silencer. Recent Research performed on the frequency domain of the selected decomposed audio subband with Singular Value Decomposition (SVD), then Cartesian to Polar Transformation (CPT) converts the top 2 singular values before inserting data using the QIM technique. Audio watermarking performance is determined by several parameters, among others durability (BER) with a maximum value of 15%-25%, audio quality objectively (SNR / ODG) with a minimum SNR value of 20 dB or ODG with a minimum value -1, subjective audio quality (MOS) with an average value of at least 4 and high insertion capacity with a value of at least 8 bits per second (bps). Besides The durability of watermarking is tested with a Stirmark benchmark to find out how many stirmark attacks can be anticipated by the techniques developed. In the realization of audio watermarking, the performance parameters are not reinforce each other, but there are trade-offs between these parameters. If the capacity of an insert method is increased it will decrease the quality and durability of watermarking, and vice versa, so it is necessary to set input parameters for each method in order to produce Insertion performance with good audio quality with strong durability against standard watermarking attacks and sufficient capacity. Research using DCT magnitude modification techniques using Fibonacci series is a new technique that produces very high capacity. Data hiding uses the Fibonacci sequence in the frequency domain of the audio using DCT provides higher insertion capacity than FFT at 500.27 bps, with excellent audio quality on ODG -0.3. MC watermarking technique in audio is a new, modulating technique a watermark with multiple carrier signals with the resulting audio quality provide ODG -2.26 with a fairly good perceptual, 40 bps capacity, and good resistance to 128 kbps MP3 attacks and resampling. In the QIM technique based on the LWT-DST-SVD-CPT domain, the novelty of the technique there is the use of DST combined with LWT, SVD and CPT. Simulation results with a watermark capacity of 172.26 bps, obtained quality 27.44-29.3 dB and resistance to MP3 attacks 128 kbps, 9 kHz LPF, resampling 22 kHz, and 20 dB AWGN. While the QIM technique based on the SWT-DST-QR-CPT domain, novelty is also found in the use of DST combined with SWT, QR, CPT. The resulting audio quality is very good in the range SNR 29.88-37.15 dB, capacity 21.43 bps, with good resistance at 64 kbps MP3 attacks, 9 kHz LPF, 22 kHz resampling and 20 dB AWGN. The SS technique with the LWT-FFT domain is a new SS technique because performed on the frequency domain with certain subband. Simulation results show very good audio quality in the range of SNR 30.2-34.99 dB, capacity 172.26 bps, with good resistance to 64 kbps MP3 attacks, 9 kHz LPF, 16 kHz resampling and 8 bit requantization. vThe multiple-bit SS technique is based on Compressive Sampling using the method SVD truncating is a new technique. This technique provides benefits by inserting and compressing the audio signal, so it can be used in the process of taking an important 1-dimensional audio signal or discrete signal but needs to be compressed because of the large number of samples received. This technique produces a range of good performance parameters with ODG -0.94 to -0.74, watermark capacity of 729-5292 bps, compression ratio 1.47× to 4.84× and resistant to noise attacks, because a 10 dB AWGN attack this technique obtains an acceptable detected watermark. The novelty of the multi-bits SS technique with SWT is the use of a stationary subband as an insertion place with a digital watermark modulated with multiple SS bits. The detection process at the recipient can also be done without having to know where the insertion of the watermark is on the embedding system. The result of this technique is more robust than the previous related technique. With an average ODG audio quality of -0.68, a digital watermark capacity of 84 bps, this technique is impervious with BER = 0 against MP3 attacks starting at 64 kbps, 8 kHz LPF, requantization, AWGN with SNR starting at 0 dB, and amplitude scaling attacks with gain 1.5× and 2.2×.