Problem and Motivation Mobile multimedia application is vulnerable to malicious attacks in wireless communication Significant computation for video encryption is expected on battery-operated mobile devices Evaluate symmetric video encryption schemes from the perspective of energy consumption both analytically and experimentally Overview of Secure Video Applications network Symmetric Encryption Compressed Bit Stream Encrypted & Raw Video Video Encoder Shuffling Coefficients Naive Selective VEA Zig-Zag Motion Estimation Quantization DCT Entropy Encoding Secure Video Encoder Decryption Decompressed Video Decoder Unshuffling Decoding Inverse IDCT Compensation Secure Video Decoder Malicious Attacks Battery -Operated Devices This is an study of video encryption for mobile handheld devices in terms of energy consumption. Emerging mobile multimedia applications such as military video conferencing and video surveillance is subject to malicious attacks during transmission, especially in wireless communication. And we expect video encryption consumes pretty high energy because of the significant computation for symmetric encryption techniques. So we evaluate several video encryption techniques in terms of energy consumption for mobile devices by analysis and experiments as well. They are typical steps of video application and in order to make it secure, they encrypt video data right after entropy encoding and decrypt it before general steps of video decoding. At first we evaluate the previous video encryption schemes analytically in terms of energy consumption. Naïve encryption scheme encrypts the whole video data with symmetric encryption technique such as DES and AES. It is highly secure but slow compared to selective encryption scheme. Selective encryption scheme encrypts partial data of video like intra-frames or intra-blocks. It’s less secure than Naïve algorithm but faster. On the other hand, Zig-Zag permutation algorithm shuffles coefficients before entropy encoding. However it’s not secure and it breaks the efficiency of video encoding. The basic idea of VEA is to use the uniform distribution of byte values so they encrypts the half of the stream after xoring two halves. But when I looked at the distribution of byte value in h.263 encoded video data, which are not distributed uniformly. So we don’t consider the last two encryption schemes and analysis shows the relative energy consumption of each video encryption scheme and full encryption, in other words naïve video encryption scheme, may consume energy as twice as partial encryption according to the portion of intra-blocks. To measure the power consumption, we implement the real video encoder and decoder with full encryption technique and partial encryption and set up the experiments as shown in this picture using sharp zaurus PDA. And we define the level of quality and security based on video quality dominated by quantization scale and degree of security using full encryption or partial encryption. As we expect, higher quality and security video encoder consumes higher energy. However the energy overhead for full video encryption is not a big deal, which shows just 2 % energy overhead when we look at the level 5 and 6, which are appropriate for mobile application. Other experimental shows that energy consumption for video encryption is negligible compared to video encoding and energy consumption of encryption is very small irrespective of video clips. Any questions? Analytical Study of Video Encryption Schemes with respect to Energy Consumption Studied Video Encryption Schemes Analytical Comparison of Video Encryption Schemes (1) Naïve Encryption Scheme Treats video data like a text stream with symmetric encryption technique (e.g. DES) Highly secure but slow No change in size after encryption Encrypts the partial video (e.g. Intra-blocks) Med. secure but faster than full NAÏVE No change in size after encryption (2) Selective Encryption Scheme Shuffles the coefficients Insecure but very fast Breaks the efficiency of encoding (3) Zig-Zag Permutation Scheme Encrypts half of the stream after XORing Secure and comparably fast But byte values of data are not uniformly distributed in case of H.263 (4) Video Encryption Algorithm (VEA) Algorithm Relative Energy Naive 100 % Selective 59 % I-frame P-frame Video Encoding & Zig-Zag Permutation (H.263 & Shuffle) Video Encoding (H.263) VEA (XOR & DES) Intra-block I-frame P-frame Selective Encryption (DES) Video Encoding (H.263) I-frame P-frame Naive Encryption (DES) Video Encoding (H.263) P-frame P-frame I-frame P-frame P-frame I-frame Zig-Zag < 1 % VEA 50 % EnergyZig-Zag = eoverhead eoverhead - energy to shuffle coefficients EnergyVEA = ½*(eDES+eXOR)*STotal eXOR - energy for XOR EnergyNaive=eDES*STotal EnergySelective = eDES*SIB eDES - energy to encrypt one byte by DES STotal - size of the whole video data Full Encryption (Naïve) may consume energy as twice as Partial (Selective) Encryption SIBl - size of Intra-blocks in video data Experimental Study on tradeoffs between Quality with Security and Energy Consumption Experimental Setup Measured Energy Consumption of Quality & Security Videos Energy overhead for full video encryption is NEGLIGIBLE Quality & Security Quality (Quant Scale) (Full vs. Partial) 1 2 3 4 5 6 7 8 High (Quant = 1) Mid-High (Quant = 4) Mid-Low (Quant = 10) Low (Quant = 31) High (Full) Low (Partial) Measured Energy (J) Energy Overhead 128.2 111.0 92.05 83.56 77.62 75.78 70.44 69.89 13 % 9 % 2 % 1 % Higher Quality of Video High Degree of Security Experimental Results Negligible Energy Overhead PZaurus = * VZaurus VR R 80 74.77 90 77.62 70 80 74.77 75.78 74.11 72.26 72.87 60 Huge Difference (98%) 70 (2.4%) (1.7%) System Architecture 50 60 Secure Video Application (Encoder / Decoder) Measured Energy (Joules) 50 Encoding without Encryption 40 Measured Energy (Joules) 5 V V Zaurus Encoding with Encryption (Selective) 40 DES H.263 Codec Device Driver 30 Encoding with Encryption (Naïve) 30 OpenSSL Library 20 R = 22 ohm 11.37 20 Operating System (LINUX) 10 1.5 10 Mobile Handheld Hardware Application V R FOREMAN.qcif NEWS.qcif 400 MHz XScale 64 MB flash 32 MB SDRAM H.263 Encoder H.263 Decoder DES Crypto 11 MB with 300 frames 1:10(IP ratio),10(Quant),full search Video Clips DAQ board with BNC Connector Windows XP 1,000 samples/sec Encryption consumes negligible Energy as compared to encoding Energy consumption of encryption is negligible irrespective of video clips Power Measurement System