hash stringlengths 32 32 | doc_id stringlengths 7 13 | section stringlengths 3 121 | content stringlengths 0 2.2M |
|---|---|---|---|
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6 TEA2 ALGORITHM DESCRIPTION | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1 TEA2 Functional Components | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.1 Summary of Components | The cryptographic algorithm TEA2 consists of the following functional components as depicted in Figure 7: β’ A set of eight 8-bit wide shift registers, called Output Register (clause 6.1.3). β’ A set of ten 8-bit wide shift registers, called Cipher Key Register (clause 6.1.4). β’ A byte permutation function P (byte substitution) (clause 6.1.5). β’ Two functions E to expand the 16-bit output of two 8-bit registers to 32 bits word (clause 6.1.6). β’ Two non-linear functions Ζ1 and Ζ2 to compute a byte from the expanded 32-bit word (clause 6.1.7). β’ An 8-bit permutation function BP (wire crossing) (clause 6.1.8). β’ Five 8-bit wide XOR functions to combine section outputs (bytes) and functions in the feedback of the Cipher key register and the output register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.2 Notation | The symbol β denotes the bytewise addition modulo two (exclusive or); i.e. msb of byte x is added to msb of byte y, ...., lsb of byte x to lsb of byte y. x (i)1 x(i)2 x(i)1 β x(i)2 0 0 1 1 0 1 0 1 0 1 1 0 x(i) is bit i of byte x y(i) is bit i of byte y i = 0..., 7 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.3 Output Register | The output register is denoted by R0, R1..., R6, and R7, and functions as an 8-bit wide shift register. The output bytes of sections R0 up to R5 are also used as input for the other functional components of TEA2. The output of R7 is added to the outputs of the other functional components and returned to the first section R0 as depicted in Figure 7. Each 19 steps the byte output of R7 is used as key byte for encryption or decryption. Before the process as described above can take place the output register is loaded with an Initialization Vector (see clause 6.2.3) and initialized (see clause 6.2.4). ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 16 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.4 Cipher Key Register | The Cipher key (CK) register, denoted by K0, K1, ....., K8, and K9, is a ten section 8-bit wide shift register. The bytes of a CK are loaded into the CK register as shown in Figure 8 and before the Initialization Vector is loaded into the Output Register. During initialization and the generation of Cipher bytes the Cipher key mixing is done, according to the irreducible polynomial X10 + X3 + 1 and the permutation function P, as follows: with K'i denoting the next byte value (i.e. after one step) of the register section Ki: K'0 = P (K9 β K2) K'1 = K0 K'2 = K1 K'3 = K2 K'4 = K3 K'5 = K4 K'6 = K5 K'7 = K6 K'8 = K7 K'9 = K8 The series of the mixed and permutated CK bytes is also offered to the feedback circuit of the output register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.5 Byte Permutation Function | The CK byte permutation function P is an arbitrary permutation on 256 bytes. The look-up table for this permutation is given in Figure 9. The input to the function is two 4-bit nibbles, one higher nibble and one lower nibble. In the output, the left 4-bit nibble is the higher one, e.g. P (37) = A1. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.6 Expander E | The expander function converts a two-byte input to a 32-bit word, i.e. eight 4-bit nibbles for the non-linear function Ζ1 and Ζ2. The structure of expander E and functions Ζ1 and Ζ2 is shown in Figure 10. In this figure, bits 1-8 are the bits of R1, respectively R4. Bits 9-16 are the bits of R0, respectively R3. Bits 1 and 9 are the most significant bits. The 4-bit nibble inputs to the Si boxes are the result of reading R1 R0 (for the input of Ζ1), respectively R4 R3 (for the input of Ζ2). The bit left is the msb of each 4-bit nibble. 6.1.7 Nonlinear Function Ζ1 The non-linear functions Ζ1 and Ζ2 have the structure shown in Figure 10. Each of the boxes S1 to S8 receives the nibble input concerned from the expander E and computes a bit for the output byte. S1 is the most significant bit in the output byte, and S8 is the least significant one. The functions Ζ1 and Ζ2 are given in the truth tables Figure 11 and Figure 12, respectively. The hexadecimal value of the input nibble for each Si is the one given in the top row of these figures. The computed output bit for the corresponding Si can be read in the column below the input nibble row. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.8 8-bit Permutation BP | The permutation BP is a so-called wire crossing, i.e. only one permutation with a fixed pattern. If the eight bits of R5 are numbered 12345678, the order of the bits after BP becomes 48572136. The left bit in both bytes is the most significant and the right bit is the least significant. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 17 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.1.9 Feedback of output register | The results of the functional components BP, Ζ1, Ζ2 and CK byte permutation P are added in the feedback path of the output register and affect the operation (i.e. operation after loading the CK and Initialization vector) as follows: β’ with R'i denoting the next byte value (i.e. after one step) of the output register section Ri: - R'0 = R7β BP(R5) β R2 β Ζ1 [E (R1, R0)] β POut R'1 = R0 - R'2 = R1 - R'3 = R2 β Ζ2 [ E (R4, R3)] - R'4 = R3 - R'5 = R4 - R'6 = R5 - R'7 = R6 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2 Key Stream Generation | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2.1 Summary | The algorithm consists of four main phases: the CK loading, the IV loading, the run-up and the key byte generation proper. During the CK loading, the initial state of the Cipher key register is determined. Next, the initial state of the output register is determined by loading of the Initialization Vector as described in clause 6.2.3. Thereafter, a run-up is carried out during which the initial contents of the Output Register (converted and adapted IV) and the CK Register (CK) are changed by the effect of the BP, E, Ζ1, Ζ2 and P functions (see clause 6.2.4). After the run-up is completed, each output cycle produces a key byte from the key byte stream as specified in clause 6.2.5. At any point during the run-up and the generation of the key byte stream, the state of the algorithm is determined by the states of the output and the CK registers. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2.2 CK loading | The CK loading depends solely on the 80-bit Cipher Key. The CK bytes are loaded as depicted in the Load Map shown in Figure 8. The CK bytes C1, C2, ....., C10 are shifted into the register from K0 to K9. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2.3 IV loading | The output register is first loaded with a 29-bit Initialization Vector. This IV is converted to a 32-bit word by taking the three most significant bits as zeroes, and the remaining 29 bits as the given IV. For instance, if the IV is the binary value: 11010 00011010 11100010 00000110 the 32-bit word becomes: 00011010 00011010 11100010 00000110. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 18 The least significant byte of that 32-bit word is loaded into R3, the next byte into R4, the next one into R5, and, finally, the most significant (the padded one) becomes the R6 initial value. The cells R0, R1, R2 and R7 are loaded in the same order with the 32-bit word XORed with the constant mask 5A6E3278. This gives the following initialization assignments, with the 32-bit word as a four-byte number F1F2F3F4: R7 = F1 β 5A R6 = F1 R5 = F2 R4 = F3 R3 = F4 R2 = F2 β 6E R1 = F3 β 32 R0 = F4 β 78 Using the IV, mentioned in the example above, this produces: 1 8 01000000 00011010 00011010 11100010 00000110 01110100 11010000 01111110 for the: R7, ..., R4, and R3, ..., R0 bytes. The eight bits of each Ri numbered from 1 to 8 are loaded into the register locations Ri (7), Ri (6), ..., Ri (0) respectively. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2.4 Run-up | After the IV load into the output register all functional components of TEA2 become operational and 50 initializing steps are performed to bring the algorithm in the CK and IV dependent starting point from which the generation of key bytes can start. For run-up and key byte generation a step is defined as applying the formulae in clauses 6.1.4 and 6.1.9 once. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.2.5 Key byte generation | After the run-up cycle one step is made to produce the first key byte. Successive key bytes are generated each 19 steps. More precisely, the keystream generator, being byte orientated, generates 8 output bits at a time. Keystream bytes are generated by stepping the algorithm 19 times and then taking the value held in the Output Register stage R7. The 8 bits in this register are then used, most-significant bit first, as the next 8 bits of the keystream. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 19 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 6.3 Figures of TEA2 Algorithm | R0....R7 Output Register (eight 8-bit wide registers) K0....K9 Cipher Key load register (ten 8-bit wide registers) E Expansion from 16 to 32 bits Ζ1, Ζ2 Non-linear function: each eight functions of 4 bits BP Permutation of 8 bits (wire crossing) P Permutation on 28 elements (byte substitution) Figure 7: Functional components of TEA2 CK = C1 / C2 / C3 / C4 / C5 / C6 / C7 / C8 / C9 / C10 with: Ci = CK byte i Cipher Key K9 K8 K7 K6 K5 K4 K3 K2 K1 K0 C1 C10 Figure 8: Cipher Key Load Map ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 20 H 0 1 2 3 4 5 6 7 8 9 A B C D E F g h 0 62 DA FD B6 BB 9C D8 2A AB 28 6E 42 E7 1C 78 9E e 1 FC CA 81 8E 32 3B B4 EF 9F 8B DB 94 0F 9A A2 96 r 2 1B 7A FF AA C5 D6 BC 24 DF 44 03 09 0B 57 90 BA 3 7F 1F CF 71 98 07 F8 A1 60 F7 52 8D E5 D7 69 87 n 4 14 ED 92 EB B3 2F E9 3D C6 50 5A A7 45 18 11 C4 i 5 CE AC F4 1D 82 54 3E 49 D5 EE 84 35 41 3A EC 34 b 6 17 E0 C9 FE E8 CB E6 AE 68 E2 6B 46 C8 47 B2 E3 b 7 97 10 0E B8 76 5B BE F5 A6 3C 8F F6 D1 AF C0 5E l 8 7E CD 7C 51 6D 74 2C 16 F2 A5 65 64 58 72 1E F1 e 9 04 A8 13 53 31 B1 20 D3 75 5F A4 56 06 8A 8C D9 A 70 12 29 61 4F 4C 15 05 D2 BD 7D 9B 99 83 2B 25 i B D0 23 48 3F B0 2E 0D 0C C7 CC B7 5C F0 BF 2D 4E n C 40 39 9D 21 37 77 73 4B 4D 5D FA DE 00 80 85 6F p D 22 91 DC 26 38 E4 4A 79 6A 67 93 F3 FB 19 A0 7B u E F9 95 89 66 B9 D4 C1 DD 63 33 E1 C3 B5 A3 C2 27 t F 0A 88 A9 1A 6C 43 EA AD 30 86 36 59 08 55 01 02 | Left = higher nibble Figure 9: Byte permutation lookup table 1 8 9 16 Figure 10: Structure of expander and functions Ζ1 and Ζ2 byte 1 byte 2 E 1 2 15 16 2 3 16 9 3 4 9 10 4 5 10 11 5 6 11 12 6 7 12 13 7 8 13 14 8 1 14 15 S1 S2 S3 S4 S5 S6 S7 S8 byte ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 21 Value of Input Nibble 0 1 2 3 4 5 6 7 8 9 A B C D E F S1 1 1 0 1 0 0 0 1 0 1 1 0 0 0 1 1 S2 0 1 1 1 0 0 0 1 1 1 0 0 0 1 1 0 S3 1 0 1 1 0 0 1 0 1 1 0 0 1 0 0 1 S4 0 0 1 0 1 0 0 1 1 1 0 0 1 1 1 0 S5 0 1 1 0 1 0 1 1 1 0 0 0 1 1 0 0 S6 0 0 0 1 0 0 1 1 0 1 1 0 1 1 0 1 S7 1 0 1 0 0 1 1 1 0 1 1 0 0 0 0 1 S8 1 0 0 1 1 1 1 0 1 0 1 0 0 1 0 0 Figure 11: Truth table of Ζ1 Value of input nibble 0 1 2 3 4 5 6 7 8 9 A B C D E F S1 1 0 0 0 1 0 1 1 0 0 1 1 0 1 1 0 S2 0 1 0 0 1 1 0 1 1 0 0 1 0 0 1 1 S3 0 0 0 1 0 1 1 1 0 1 1 0 1 1 0 0 S4 1 0 0 0 1 1 1 0 0 0 1 1 1 0 0 1 S5 0 1 1 1 1 0 0 1 1 1 0 0 0 1 0 0 S6 1 0 0 1 0 0 1 1 0 1 0 0 1 1 0 1 S7 1 0 0 0 0 1 0 1 1 1 1 0 1 0 0 1 S8 0 1 0 1 0 0 0 1 0 1 1 0 1 0 1 1 Figure 12: Truth table of Ζ2 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7 TEA3 ALGORITHM DESCRIPTION | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1 TEA3 Functional Components | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.1 Summary of Components | The cryptographic algorithm TEA3 consists of the following functional components as depicted in Figure 13: β’ An Output Register comprising a set of eight 8-bit wide shift register stages (clause 7.1.3). β’ A Cipher Key Register comprising a set of ten 8-bit wide shift register stages (clause 7.1.4). β’ A byte permutation function P (byte substitution) (clause 7.1.5). β’ Two instances of a function E which expands the 16 bits formed from the concatenation of two 8-bit registers to 32 bits (clause 7.1.6). β’ Two nonlinear functions Ζ1 and Ζ2 which each compute one byte from the 32-bit output of the expander function E (clause 7.1.7). β’ An 8-bit permutation function BP (wire crossing) (clause 7.1.8). β’ Five 8-bit wide XOR functions to combine register stage outputs and function outputs in the feedback of the Cipher Key Register and the Output Register. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 22 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.2 Notation | The symbol β denotes the bytewise addition modulo two (exclusive or); i.e. msb of byte x is added to msb of byte y, ...., lsb of byte x to lsb of byte y. x (i)1 x(i)2 x(i)1 β x(i)2 0 0 0 0 1 1 1 0 1 1 1 0 x(i) is bit i of byte x y(i) is bit i of byte y i=0β¦..,7 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.3 Output Register | The Output Register comprises eight stages denoted by bytes R0 .....R7 and functions as an eight stage 8-bit wide shift register. Stages R1 and R2, and R4, R5 and R6 are also used as input for the other functional components of TEA3. Stage R7 is added to the outputs of the other functional components and returned to the first stage R0 as defined below and as depicted in Figure 13. The outputs of the functional components BP, E, Ζ1, Ζ2, and the Cipher Key Register (Kout) are added in the feedback paths of the Output Register and define its operation as follows: R'0 = R7 β BP [R4] β Ζ2 (E [R2,, R1]) β Kout R'1 = Ro R'2 = R1 R'3 = R2 R'4 = R3 R'5 = R4 β Ζ1 (E [R6, R5]) R'6 = R5 R'7 = R6 with R'i denoting the next byte value (i.e. after one step) of the Output Register stage Ri, and Kout denoting the output from the Cipher Key Register block (i.e. Kout = K9 β P [K7 β K2]). Before the process as described above can take place the Output Register is loaded with an Initialization Vector (see clause 7.2.3). Following loading it is initialized (see clause 7.2.4). Once initialized, the algorithm is repeatedly stepped 19 times and the resultant contents of byte R7 following each 19/h step is used as a keystream byte for encryption or decryption. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.4 Cipher Key Register | The Cipher Key Register comprises ten stages denoted by bytes K0 ......K9 and functions as a ten stage 8-bit wide shift register. Stages K2 and K7 are also used as input for the other functional components of TEA3. Stage K9 is added to the outputs of the other functional components and returned to the first stage K0 as defined below and as depicted in Figure 13. During run-up and the generation of keystream bytes the Cipher Key Register is re-circulated as follows (P is the permutation function): K'0 = Kout = K9 β P (K7 β K2) K'1 = K0 ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 23 K'2 = K1 K'3 = K2 K'4 = K3 K'5 = K4 K'6 = K5 K'7 = K6 K'8 = K7 K'9 = K8 with K'i denoting the next byte value (i.e. after one step) of the register section Ki, Kout is an input to the feedback circuit of the Output Register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.5 Byte Permutation Function P | The byte permutation function P is a fixed random-like permutation on 256 bytes. The look-up table for this permutation is given in Figure 15. The input to the function is two 4-bit nibbles, one higher nibble and one lower nibble. In the output, the left 4-bit nibble is the higher one, e.g. P (27) = 5B. In this example '2' is the higher input nibble, '7' is the lower input nibble, '5' is the higher output nibble and 'B' is the lower output nibble. The example is indicated by the shaded cell in Figure 15. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.6 Expander E | The expander function converts a two-byte input to a 32-bit word. The 32-bits are subdivided into eight 4-bit nibbles for the nonlinear functions Ζ1 and Ζ2. The structure of expander E and functions Ζ1 and Ζ2 is shown in Figure 16. In this figure, bits 1-8 are the bits of R6 (R2). Bits 9-16 are the bits of R5 (R1). Bits 1 and 9 are the most significant bits. The 4-bit nibble inputs to the Si boxes are the result of reading R6 R5 (for the input of Ζ1), and R2 R1 (for the input of Ζ2). The left bit is the msb of each 4-bit nibble. 7.1.7 Nonlinear Function Ζ1 The nonlinear functions Ζ1 and Ζ2 have the structure shown in Figure 16. Each of the boxes S1 to S8 receives the nibble input concerned from the expander E and computes a bit for the output byte. S1 is the most significant bit in the output byte, and S8 is the least significant one. The functions Ζ1 and Ζ2, are given in Figure 17 and Figure 18, respectively. The hexadecimal value of the input nibble for each Si is the one given in the top row of these figures. The computed output bit for the corresponding Si can be read in the column below. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.1.8 8-bit Permutation BP | The permutation BP is a simple wire-crossing, i.e. a reordering of the bits within the byte. If the eight bits of R4 are numbered 12345678, the order of the bits after BP becomes 38467215. The left bit (bit 1) in both bytes is the most significant and the right bit (bit 8) is the least significant. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2 Keystream Generation | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2.1 Summary | The algorithm consists of three main phases: the CK and the IV loading, the run-up and the keystream generation proper. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 24 During the loading phase the Cipher Key is used to set the initial state of the Cipher Key Register (clause 7.2.2), and the Initialization Vector is used to set the initial state of the Output Register (clause 7.2.3). Following the initializations of the two registers a run-up phase is carried out to complete the initializations of the algorithm (see clause 7.2.4). After the run-up is completed, each output cycle produces 8 bits of keystream as specified in clause 7.2.5. At any point during the run-up and the generation of the keystream, the state of the algorithm is determined by the states of the Output Register and the Cipher Key Register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2.2 CK loading | The 80-bit Cipher Key is used to initialize the ten stage Cipher Key Register as depicted in the Load Map shown in Figure 14. The CK bytes C1, C2β¦, 10, starting with the most significant byte (C1), are shifted into the Cipher Key Register Ki, entering through stage K0. Note that the feedback function is not enabled at this time. I.e. the bytes of CK are initially shifted into the register such that: K10-i = Ci, i = 1β¦10 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2.3 IV loading | The 29-bit IV is converted to a 32-bit word (four bytes) by adding three '0' bits to the most significant end. For instance, if the IV is the binary value: 11010 00011010 11100010 00000110 the 32-bit word becomes: 00011010 00011010 11100010 00000110. The resultant 32-bit IV is used to initialize the Output Register R. The least significant byte of that 32-bit word is loaded into R3, the next byte into R4, the next one into R5, and, finally, the most significant into R6. The stages R0, R1, R2 and R7 are loaded in the same order with the 32-bit word XORed with the hexadecimal constant C43A7D51. So, taking the (adjusted) IV as IV1IV2IV3IV4, we have the following initializations of R: R7 = IV1 β C4 R6 = IV1 R5 = IV2 R4 = IV3 R3 = IV4 R2 =IV2 β 3A R1 = IV3 β 7D R0 = IV4 β 51 Using the IV, mentioned in the example above, we get: 1 8 11011110 00011010 00011010 11100010 00000110 00100000 10011111 01010111 for the R7. ... .. R0 bytes. The eight bits of each Ri numbered from 1 to 8 are loaded into the register locations Ri (7), Ri (6), ..., Ri (0) respectively. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 25 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2.4 Run-up | After the CV and IV have been loaded into their respective registers all functional components of the TEA3 become operational and 32 steps are performed to bring the algorithm to a state where the generation of keystream can start. For run-up and keystream generation a step is defined as applying the formulae in clause 7.1.4 once. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.2.5 Key byte generation | The keystream generator, being byte orientated, generates 8 output bits at a time. Keystream bytes are generated by stepping the algorithm 19 times and then taking the value held in the Output Register stage R7. The 8 bits in this register are then used, most-significant bit first, as the next 8 bits of the keystream. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 7.3 Figures of TEA3 Algorithm | Figure 13: Functional components of TEA3 ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 26 Figure 14: Cipher Key Load Map Figure 15: Byte permutation lookup table ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 27 Figure 16: Structure of expander and functions Ζ1 and Ζ2 Figure 17: Truth table of Ζ1 Figure 18: Truth table of Ζ2 ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 28 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8 TEA4 ALGORITHM DESCRIPTION | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1 TEA4 Functional Components | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.1 Summary of Components | The cryptographic algorithm TEA4 consists of the following functional components as depicted in Figure 19: β’ An Output Register comprising a set of eight 8-bit wide shift register stages (clause 8.1.3). β’ A Cipher Key Register comprising a set of seven 8-bit wide shift register stages (clause 8.1.4). β’ A byte permutation function P (byte substitution) (clause 8.1.5). β’ Two instances of a function E which expands the 16 bits formed from the concatenation of two 8-bit registers to 32-bits (clause 8.1.6). β’ Two nonlinear functions Ζ1 and Ζ2 which each compute one byte from the 32-bit output of the expander function E (clause 8.1.7). β’ An 8-bit permutation function BP (wire crossing) (clause 8.1.8). β’ Six 8-bit wide XOR functions to combine register stage outputs and function outputs in the feedback of the Cipher Key Register and the Output Register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.2 Notation | The symbol β denotes the bytewise addition modulo two (exclusive or); i.e. msb of byte x is added to msb of byte y, ...., lsb of byte x to lsb of byte y. x (i)1 x(i)2 x(i)1 β x(i)2 0 0 0 0 1 1 1 0 1 1 1 0 x(i) is bit i of byte x y(i) is bit i of byte y i=0β¦..,7 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.3 Output Register | The Output Register comprises eight stages denoted by bytes R0 .....R7 and functions as an eight stage 8-bit wide shift register. Stages R1 and R2, and R4, R5 and R6 are also used as input for the other functional components of TEA4. Stage R7 is added to the outputs of the other functional components and returned to the first stage R0 as defined below and as depicted in Figure 19. The outputs of the functional components BP, E, Ζ1, Ζ2, and the Cipher Key Register (Kout) are added in the feedback paths of the Output Register and define its operation as follows: R'0 = R7 β BP[R6] βΖ1 (E [R5, R4]) β Kout R'1 = R0 R'2 = R1 R'3 = R2 R'4 = R3 β Ζ2 (E [R1, R0]) ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 29 R'5 = R4 R'6 = R5 R'7 = R6 with R'i denoting the next byte value (i.e. after one step) of the Output Register stage Ri, and Kout denoting the output from the Cipher Key Register block (i.e. Kout = K6 β P (K5 β K1)). Before the process as described above can take place the Output Register is loaded with an Initialization Vector (see clause 3.3). Following loading it is initialized (see clause 8.2.4). Once initialized, the algorithm is repeatedly stepped 19 times and the resultant contents of byte R7 following each 19th step is used as a keystream byte for encryption or decryption. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.4 Cipher Key Register | The Cipher Key Register comprises seven stages denoted by bytes Koβ¦.K6 and function as a seven stage 8-bit wide shift register. Stages K1 and K5 are also used as input for the other functional components of TEA4. Stage K6 is added to the outputs of the other functional components and returned to the first stage Ko as defined below and as depicted in Figure 19. During run-up and the generation of keystream bytes the Cipher Key Register is re-circulated as follows (P is the permutation function): K'0 = Kout = K6 β P (K5 β K1) K'1 = K0 K'2 = K1 K'3 = K2 K'4 = K3 K'5 = K4 K'6 = K5 with K'i denoting the next byte value (i.e. after one step) of the register section Ki, Kout is an input to the feedback circuit of the Output Register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.5 Byte Permutation Function P | The byte permutation function P is a fixed random-like permutation on 256 bytes. The look up table for this permutation is given in Figure 21. The inputs to the function are two 4-bit nibbles, one higher nibble and one lower nibble. In the output, the left 4-bit nibble is the higher one, e.g. P (28) = D4. In this example '2' is the higher input nibble, '8' is the lower input nibble, 'D' is the higher output nibble and '4' is the lower output nibble. The example is indicated by the shaded cell in Figure 21. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.6 Expander E | The expander function conve1ts a two-byte input to a 32-bit word. The 32-bits are subdivided into eight 4-bit nibbles for the nonlinear function Ζ1 and Ζ2. The structure of expander E and functions Ζ1 and Ζ2 is shown in Figure 22. In Figure 22, bits 1-8 are the bits of R5 (R1). Bits 9-16 are the bits of R4 (R0). Bits 1 and 9 are the most significant bits. The 4-bit nibble inputs to the Si boxes are the result of reading R5 R4 (for the input of Ζ1), and R1 R0 (for the input of Ζ2). The left bit is the msb of each 4-bit nibble. 8.1.7 Nonlinear Function Ζ1 The nonlinear functions Ζ1 and Ζ2 have the structure shown in Figure 22. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 30 Each of the boxes S1 to S8 receives the nibble input concerned from the expander E and computes a bit for the output byte. S1 is the most significant bit in the output byte, and S8 is the least significant one. The functions Ζ1 and Ζ2, are given in truth tables, Figure 23 and Figure 24, respectively. The hexadecimal value of the input nibble for each Si is the one given in the top row of these figures. The computed output bit for the corresponding Si can be read in the column below the input nibble row. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.1.8 8-bit Permutation BP | The permutation BP is a simple wire-crossing, i.e. a reordering of the bits within the byte. If the eight bits of R4 are numbered 12345678, the order of the bits after BP becomes 74638152. The left bit (bit 1) in both bytes is the most significant and the right bit (bit 8) is the least significant. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2 KEYSTREAM GENERATION | |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2.1 Summary | The algorithm consists of three main phases: the CK and the IV loading, the run-up and the keystream generation proper. During the loading phase the Cipher Key is used to set the initial state of the Cipher Key Register (clause 8.2.2), and the Initialization Vector is used to set the initial state of the Output Register (clause 8.2.3). Following the initializations of the two registers a run-up phase is carried out to complete the initializations of the algorithm (see clause 8.2.4). After the run-up is completed, each output cycle produces 8 bits of keystream as specified in clause 8.2.5. At any point during the run-up and the generation of the keystream, the state of the algorithm is determined by the states of the Output Register and the Cipher Key Register. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2.2 CK loading | The 80-bit Cipher Key is used to initialize the seven stage Cipher Key Register as defined below and as depicted in Figure 20. All stages of the Cipher Key Register K are initially set to 0, and then the Cipher Key bytes are mixed into the register starting with the most significant byte of the key (C1). The feedback defined in clause 8.1.4 is inoperative at this time. The bytes of the Cipher Key (CK) are loaded into the key register in the following way: Step 1: K0 = K 1 = K2 = K3 = K4 = Ks= K6 = 0 {Set all stages to 0} Step 2: {Load the Cipher Key bytes} For i = 1 To 10 Do K'0 = CKi β K6 β P (K5 β K1) K'1 = K0 K'2 = K1 K'3 = K2 K'4 = K3 K'5 = K4 K'6 = K5 where K'i is denoting the next byte value (i.e. after one step) of the register section Ki. I.e. During loading the CKi bytes are added modulo two to K6 and to the result of substitution by the permutation function P of the modulo two addition of the output of Sections K1 and K5. This result is then passed to section K0. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 31 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2.3 IV loading | The 29-bit IV is converted to a 32-bit word (four bytes) by adding three '0' bits to the most significant end. For instance, if the IV is the binary value: 11010 00011010 11100010 00000110 the 32-bit word becomes: 00011010 00011010 11100010 00000110. The resultant 32-bit IV is used to initialize the Output Register R. The least significant byte of that 32-bit word is loaded into R3, the next byte into R4, the next one into R5, and the most significant into R6. The stages R0, R1, R2 and R7 are loaded in the same order with the 32-bit word XORed with the hexadecimal constant 56DC28E3. So, taking the (adjusted) IV as IV1IV2IV3IV4 gives the following initializations of R: Using the IV, mentioned in the example above, gives: 1 8 01001100 0001 1010 00011010 11100010 00000110 11000110 11001010 11100101 for the R7 .... R0 bytes. The eight bits of each Ri numbered from 1 to 8 are loaded into the register locations Ri (7), Ri (6), ..., Ri (0) respectively. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2.4 Run-up | After the CV and IV have been loaded into their respective registers all functional components of TEA4 become operational and 35 steps are performed to bring the algorithm to a state where the generation of keystream can start. For run-up and keystream generation a step is defined as applying the formulae in clause 8.1.4 once. |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.2.5 Keystream generation | The keystream generator, being byte orientated, generates 8 output bits at a time. Keystream bytes are generated by stepping the algorithm 19 times and then taking the value held in the Output Register stage R7. The 8 bits in this register are then used, most-significant bit first, as the next 8 bits of the keystream. ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 32 |
b9577ffd31d53882333a2ffba47a2f9e | 104 053-1 | 8.3 Figures of TEA4 Algorithm | Figure 19: Functional components of TEA4 Figure 20: Cipher Key Load Map ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 33 Figure 21: Byte permutation lookup table Figure 22: Structure of expander and functions Ζ1 and Ζ2 ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 34 Figure 23: Truth table of Ζ1 Figure 24: Truth table of f2 ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 35 Annex A (informative): Bibliography β’ ETSI TS 101 053-2: "Rules for the management of the TETRA standard encryption algorithms; Part 2: TEA2". β’ ETSI TS 101 053-3: "Rules for the management of the TETRA standard encryption algorithm TEA3". β’ ETSI TS 101 053-4: "Rules for the management of the TETRA standard encryption algorithms; Part 4: TEA4". ETSI ETSI TS 104 053-1 V1.2.1 (2025-02) 36 History Document history V1.1.1 July 2024 Publication V1.2.1 February 2025 Publication |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 1 Scope | The present document specifies DASH-IF Forensic A/B Watermarking. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 2 References | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found at https://docbox.etsi.org/Reference. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long-term validity. The following referenced documents are necessary for the application of the present document. [1] ISO/IEC 23009-1:2022: "Information technology -- Dynamic adaptive streaming over HTTP (DASH) -- Part 1: Media presentation description and segment formats". [2] ISO/IEC 13818-1:2019: "Information technology -- Generic coding of moving pictures and associated audio information -- Part 1: Systems". [3] IETF Internet Draft draft-pantos-hls-rfc8216bis-12: "HTTP Live Streaming 2nd Edition", R. Pantos. [4] IETF RFC 8949: "Concise Binary Object Representation (CBOR)", C. Bormann, P. Hoffman, December 2020. [5] IETF RFC 8610: Concise Data Definition Language (CDDL): A Notational Convention to Express Concise Binary Object Representation (CBOR) and JSON Data Structures", H. Birkholz, C. Vigano, C. Bormann, June 2019.". [6] IETF RFC 8392: "CBOR Web Token (CWT)", M. Jones, E. Wahlstroem, S. Erdtman, H. Tschofenig. May 2018. [7] IETF RFC 4648: "The Base16, Base32, and Base64 Data Encodings", S. Josefsson, October 2006. [8] UHD Forum: "Watermarking API for Encoder Integration, version 1.0.1", March 2021. [9] IEEE Std 1003.1β’ 2018 Edition, The Open Group Base Specifications Issue 7, 31 January 2018. [10] DASH-IF registry of watermarking technology vendors IDs. [11] IETF RFC 9053: "CBOR Object Signing and Encryption (COSE): Initial Algorithms", J. Schaad, August 2022. [12] IANA: "CBOR Web Token (CWT) Claims". |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 8 The following referenced documents are not necessary for the application of the present document but they assist the user with regard to a particular subject area. [i.1] DASH-IF Live Media Ingest Protocol. [i.2] Web Sequence Diagram. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 3 Definition of terms, symbols and abbreviations | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 3.1 Terms | For the purposes of the present document, the following terms apply: client-driven watermarking: action of watermarking content when the user device is performing some actions allowing it to make unique requests for content NOTE: The user device embeds a watermarking agent that is integrated with the application. client-side watermarking: action of watermarking when the user device is the sole responsible for doing the actual watermarking of content NOTE: The user device embeds a watermarking agent that is integrated with the audio-visual rendering engine. server-driven watermarking: action of watermarking content when the user device is not performing any other operation than conveying information such as tokens, between servers that are responsible for doing the actual watermarking of content that is delivered to the user device sequencing: action of returning a Variant of a segment when it is requested, based on a watermark token NOTE: Typically, this action is performed on a CDN edge server and is thus referred to as "edge sequencing". variant: alternative representation of a given segment of a multimedia asset NOTE: Typically, a Variant is a pre-watermarked version of the segment. WaterMark (WM) pattern: series of A/B decisions for every segment that is unique per user device |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 3.2 Symbols | Void. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 3.3 Abbreviations | For the purposes of the present document, the following abbreviations apply: ABR Adaptive Bit Rate AES Advanced Encryption Standard AF Adaptation Field API Application Programming Interface AVC Advanced Video Codec CBOR Concise Binary Object Representation CDDL Concise Data Definition Language CDN Content Delivery Network CMAF Common Media Application Format COSE CBOR Object Signing and Encryption CPU Central Processing Unit CWT CBOR Web Token DAI Dynamic Ad Insertion DASH Dynamic Adaptive Streaming over HTTP DRM Digital Rights Management ETSI ETSI TS 104 002 V1.1.1 (2023-08) 9 ECDH Elliptic Curve Diffie-Hellman HEVC High Efficiency Video Coding HLS HTTP Live Streaming HMAC keyed-Hashing for Message AuthentiCation HTTP Hypertext Transfer Protocol IANA Internet Assigned Numbers Authority IOP InterOPerability IP Internet Protocol ISOBMFF ISO Base Media File Format JITP Just In Time Packager JSON JavaScript Object Notation JWT JSON Web Token MPD Media Presentation Description NAL Network Abstraction Layer OTT Over The Top RIST Reliable Internet Stream Transport RTMP Real-Time Messaging Protocol RTP Real Time Protocol SEI Supplemental Enhancement Information SRT Secure Reliable Transport TS Transport Stream TV TeleVision UDP User Datagram Protocol UHD Ultra-High Definition URI Uniform Resource Identifier URL Uniform Resource Locator UUID Universally Unique IDentifier VOD Video On Demand WM WaterMark WMID WaterMark IDentifier WMT WaterMark Token |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 4 OTT Watermarking Using Variants | The objective of forensic watermarking is to deliver a unique version of a media asset to the different users consuming the asset. This is somewhat in opposition with media delivery mechanisms that aim at delivering the same asset to all users for efficiency purposes. As a result, in the broadcast era, a typical approach was to perform the watermarking operation at the very last step of the media delivery pipeline, within the end user device e.g. a set-top box. This solution has the virtue of leaving the whole media delivery pipeline unaltered but raises security and interoperability challenges when a large variety of devices owned and operated by the end user shall be supported. This is for instance the case with Over The Top (OTT) media delivery where content is consumed on mobile phones, tablets, laptops, connected TVs, etc. As a result, new forensic watermarking solutions have gained momentum that do not perform security- sensitive and complex operations in the end user realm. While such approaches require minimal changes in the end-user devices, they do mandate the media delivery pipeline to be modified accordingly. A notable example of such network-side watermarking solutions is OTT watermarking using Variants for Adaptive Bit Rate (ABR) content. In this case, the content is delivered by segments. The baseline idea is then to generate pre-watermarked Variants of each segment and to modify the delivery protocol so that each end user receives a unique sequence of Variants depending on a watermark pattern that has been assigned to the end user. The semantic of this pattern is context dependent and can be, for instance, a device identifier, an account identifier, a session identifier, etc. Figure 1 illustrates a particular case of this strategy, coined as A/B watermarking, where there are two Variants generated for each segment, each Variant containing a watermark that either encodes the information '0' or '1'. As a result, the watermarking system will require the transmission of a sequence of Variants as long as the length of the pattern to successfully recover the whole unique identifier. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 10 Figure 1: A/B watermarking concept with (a) ABR content delivery and (b) A/B Variants delivery When using Variants, the serialization process essentially boils down to delivering a unique sequence of Variants to each individual end user. There are two main families of methods to achieve this: 1) Server-driven methods, wherein the client does perform no operation related to watermarking. It simply fetches and forwards a token to the CDN that is responsible for delivering a unique sequence of Variants. 2) Client-driven methods, wherein the client is responsible for the serialization operation. For instance, it relies on some session-based digital object to tamper the URI ABR segments and thereby directly query a unique sequence of Variants from the CDN. The present document is describing the server-driven methods. Client-driven methods are not part of the present document. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5 Server-Driven Architecture and Workflows | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.1 Introduction | In the server-driven architecture, the device is unaware that content it consumes is watermarked. The device only exchanges a token with servers allowing these servers, usually CDN edges, to make the decision on which A or B Variant it delivers to the device. In the present document, an end-to-end system is presented. It includes the definition of watermarking metadata that limits the need for naming conventions by allowing the encoder to send this metadata all the way to the edge through origins to enable the sequencing of bits. The following goes through the functional architecture and describes the workflows when preparing content and when consuming content. In the following, it is assumed that the edge is a CDN edge. There are optional architectures, but this does impact the overall functional architecture and workflows. It is also assumed that multi-track content (audio and video multiplexed in one segment) is out of the scope of the present document. In addition, all the workflows are only examples of possible implementations. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.2 Functional Architecture | Figure 2 shows the simplified high-level functional architecture and the different interaction between the components that are involved in the flows when a device consumes watermarked content. Note that this also shows that content is encrypted, as watermarking will likely be added for premium content that is also encrypted and protected by a DRM system. Original asset ABR segments at different bitrates Sequence of ABR segments received by three users (a) Original asset A/B Variants of ABR segments at different bitrates Unique sequence of A/B Variants received by three users (b) Ingest Deliver Alice Bob Charlie Alice Bob Charlie Ingest Deliver ETSI ETSI TS 104 002 V1.1.1 (2023-08) 11 Figure 2: Functional architecture To consume content, a device needs, at minimum, to have an authorization token (for getting a DRM license) and a WM token that contains a WM pattern, a series of A or B decisions. The device is responsible for obtaining the required data before requesting segments to the CDN. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.3 System Configuration | Enabling or disabling the edge sequencing logic is set through the configuration to the edge. As an example, this can be useful for a service of live sporting events where only premium events require watermarking enforcement. Other moments of the day do not require it. In this case, content is still watermarked but the edge is only configured to sequence during the limited period of time of the premium event. When sequencing is disabled, the edge shall consume segments on the endpoint for Variant A. If this endpoint is not working properly, the origin shall deliver any available Variant on this endpoint. NOTE 1: When enabling watermarking, all devices that do not have a WM token will receive an error when requesting content, hence they are then forced to request such token. NOTE 2: As an example, enabling and disabling sequencing can be done with an API enable (true/false). Watermarked objects names shall include a pattern that the CDN can match to differentiate these objects from non-watermarked objects (initialization segments, subtitles, trickplay images). As an example, for a DASH manifest located at https://edge.hostname/path/to/endpoint/index.mpd that references video segments as: <SegmentTemplate timescale="60000" media="video_segment_$RepresentationID$_$Time$.mp4" initialization="video_init_$RepresentationID$.mp4" startNumber="10967120" presentationTimeOffset="903486496960"> The pattern for the differentiation of these objects from non-watermarked objects is video_segment_. One of the following identification schemes, referred as variantId in the present document, shall be used for the identification of the Variants: - A lower-case letter beginning with 'a'. Variants are then 'a', 'b' and so on. - A number beginning with 0. Variants are then 0, 1 and so on. When addressing content, variantId shall be translated into variantPath as follows: - variantPath = ${variantId} followed by '/' with the exception, that if ${variantId} is 'a' or '0' then ${variantPath} may be empty. Device Edge Encoder/ Watermarker Packager Origin A and B Variants A and B Variants A and B Variants WMT Generator DRM Server Authorization Server WM tokens ; A or B Variant WM token Content keys & DRM information Authz token ; License Authz token ETSI ETSI TS 104 002 V1.1.1 (2023-08) 12 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.4 WM Token | A WM token provides a WM pattern which is unique (for example per streaming session or per user). This pattern allows the sequencing of A/B Variants. Two tokenization schemes are defined in the present document. The first, named direct, embeds the WM pattern in the token and can be opened and interpreted by an edge irrespective of the underlying WM technology and provider. The second, named indirect, requires integration of a WM technology provider's edge sequencing software at the edge. The following are requirements on the WM token: - The token shall be a CWT token, the basic structural requirements are defined in IETF RFC 8392 [6]. - The token shall be with integer keys in "deterministically encoded CBOR" as specified in IETF RFC 8949 [4], clause 4.2. - Recipients shall process claims listed in IETF RFC 8392 [6], clause 3.1 when they are present. exp and iat shall be present. - The token shall include either a WM pattern (direct mode) or data for deriving the WM pattern (indirect mode). Absence of a wmpattern claim implies that the token is in indirect mode. - Recipients shall support direct mode and may support indirect mode. - The token shall be signed as described in clause 7 of IETF RFC 8392 [6]. Recipients shall support the HMAC 256/256 (kty number 5) and ES256 (kty number -7) algorithms. - The token shall be base64url-encoded as described in clause 5 of IETF RFC 4648 [7]. The following claims are defined and Table 1 provides the integer claim keys: wmtoken = { wmver-label ^ => wmver-value, wmvnd-label ^ => wmvnd-value, wmpatlen-label ^ => wmpatlen-value, ? wmsegduration-label ^ => wmsegduration-value, wmtoken-direct // wmtoken-indirect, * wmext-label => any } wmver-value = uint .size 1 wmvnd-value = uint .size 1 wmpatlen-value = uint .size 2 wmsegduration-value = [(wmtimeticks : uint, wmtimescale : uint)] wmext-label = int ; direct mode wmtoken-direct = { wmpattern-label ^ => wmpattern-value } wmpattern-value = COSE_Encrypt0 // COSE_Encrypt // bytes ; indirect mode wmtoken-indirect = { wmid-label ^ => wmid-value wmopid-label ^ => wmopid-value wmkeyver-label ^ => wmkeyver-value } wmid-value = text wmopid-value = uint wmkeyver-value = uint ETSI ETSI TS 104 002 V1.1.1 (2023-08) 13 Table 1: Integer Claim key values for the WM token Claim label Integer key wmver-label 300 wmvnd-label 301 wmpatlen-label 302 wmsegduration-label 303 wmpattern-label 304 wmid-label 305 wmopid-label 306 wmkeyver-label 307 wmver The version of the WM Token. Recipients shall support this claim. The present document describes version 1. wmvnd The WM technology vendor. Recipients shall support this claim. This provides the context for the key material needed for signature verification. In the direct mode, it also provides the context for the key material needed for decrypting wmpattern if needed. In the indirect mode, it identifies the vendor specific core to use. A list of WM technology vendor identifiers is available at [10]. wmpatlen The length in bits of the WM pattern. Recipients shall support this claim. wmpattern The WM pattern. Recipients shall support this claim in direct mode. It is recommended to encrypt the pattern. Recipients shall support ECDH-SS+A128KW (key type -32) as defined in IETF RFC 9053 [11]. wmsegduration The nominal duration of a segment. This claim is optional. Recipients may support this claim. When WMPaceInfo data is not available, this may allow the edge to define the index to be considered in the WM pattern. If WMPaceInfo is available, this claim shall be ignored. The array contains exactly 2 values. The first value is a duration in time ticks where its base unit is defined by the second value. The second value is the scale in number of time ticks per second. As an example, [60'000, 10'000] means that the segments are 60'000 ticks long while the scale is 10'000 ticks per second, wmsegduration is then equal to 6 seconds. wmid Used as input to derive the WM pattern for indirect mode. Recipients shall support this claim in indirect mode. The derivation algorithm is not defined in the present document and is vendor specific. wmopid Used as additional input to derive the WM pattern for indirect mode. Recipients shall support this claim in indirect mode. wmkeyver The key to use for derivation of the WM pattern in indirect mode. Recipients shall support this claim in indirect mode. Once the WM pattern is obtained from the token (either directly, decrypted or calculated), the CDN edge shall enforce big-endian convention to address a single bit in it when using the value of position (defined in clause 5.5.2). The following is an example with a WM pattern equal to 0x0A0B0C0D. Byte 0 1 2 3 bit offset 01234567 01234567 01234567 01234567 binary 00001010 00001011 00001100 00001101 hex 0A 0B 0C 0D For a value of position equal to 3, the bit to consider is highlighted in green (equal to 0). This is not any other bit, especially, those highlighted in red. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 14 For the indirect mode, there is a vendor specific core (identified by wmvnd). It is recommended that, performance-wise and software-stack-wise, it is comparable with the direct case. In other words, the vendors specific core should be based on the crypto operations which are used in the direct mode, and its performance should be equivalent. For example, the direct mode relies on one decryption operation when wmpattern is encrypted, the vendor specific core should be consisting of the similar operations to preserve the quantity of operations comparable between these two modes. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5 WMPaceInfo | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.1 Introduction | When a device requests a segment, the edge sequencing logic needs to know which bit in the unique WM pattern to consider for retrieving either A or B Variant of the requested segment before delivering it to the device. WMPaceInfo contains this mapping in addition to some data needed for content preparation. It is transmitted from the encoder (that is combined with the watermarking pre-processor) to the following servers that may need it (packager, origin, or edge). |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.2 WMPaceInfo Data | WMPaceInfo is as shown in Table 2. Table 2: WMPaceInfo data Attribute Producer Consumers Purpose variant Encoder Edge Integration, debugging position Encoder Edge Bit position in the WM pattern firstpart Encoder Packager, Origin Egress packaging lastpart Encoder Packager, Origin Egress packaging Where - variant gives the Variant identification, 0, 1 and so on. This information can be useful up to the edge for verifying that the right Variant has been obtained. - position is the index in the WM pattern to consider for this segment. Positions are zero-based. When it is equal to -1, the corresponding segment is not watermarked. For example, position=33 indicates that this segment refers to position 34 of the WM pattern. - firstpart informs whether this segment is the first one with this position value. It is equal to true if this is the case, otherwise it is equal to false. See clause 5.6.2 for further details. - lastpart informs whether this segment is the last one with this position value. It is equal to true if this is the case, otherwise it is equal to false. See clause 5.6.2 for further details. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3 Conveying WMPaceInfo | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.1 Introduction | WMPaceInfo is delivered from the encoder to other servers. There is no unique mechanism for this. The present document does not recommend one preferred option applicable for all protocols, Table 3 only presents some possible options for conveying WMPaceInfo with a preferred option for some protocols (in bold in the table). The following goes through these different options. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 15 Table 3: Possible options for conveying WMPaceInfo information Ingest protocol WMPaceInfo delivery options RTMP SEI RTP/UDP/RIST/SRT SEI, TS adaptation field HLS/TS over HTTP POST HTTP header, SEI CMAF-based protocols/formats (HLS/fMP4, DASH) over HTTP POST HTTP header, ISOBMFF box, SEI File access protocol ISOBMFF box, SEI, sidecar file |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.2 Sidecar File | When segments (discrete files or byteranges) are delivered with a file transfer protocol, it may be convenient to have WMPaceInfo data in a sidecar file. For efficiency, the WMPaceInfo data is not copied directly as some would be included multiple times. The sidecar file is of the following format (using CDDL representation in IETF RFC 8610 [5]), following recommendation of clause 5 of IETF RFC 8949 [4] and shall be encoded using deterministically encoded CBOR as specified in IETF RFC 8949 [4], clause 4.2 with integer keys. ;---------------------------------------+ ; Maps Integer Keys version = 1 segments = 2 fileSize = 3 startRange = 4 segmentRegex = 5 position = 6 firstpart = 7 lastpart = 8 ;---------------------------------------+ discrete-segment = { ?segmentRegex : text, position : int .size 2 .ge -1, ?firstpart : bool, ?lastpart : bool } byterange-segment = { startRange : uint .size 8, position : int .size 2 .ge -1 } sidecar-discrete = { version : uint .size 1, segments : [+ discrete-segment] } sidecar-byterange = { version : uint .size 1, fileSize : uint .size 8, segments : [+ byterange-segment] } sidecar = (sidecar-byterange // sidecar-discrete) When segments are discrete files: - sidecar shall contain only sidecar-discrete elements. - version is set to 1 for sidecar files compliant to the present document. - segmentRegex is a POSIX extended regular expression as described in clause 9 of [9]. It allows to define the filename of the segments for which the data applies. segmentRegex is optional. - position, firstpart and lastpart are defined in clause 5.5.2. firstpart and lastpart are optional. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 16 NOTE 1: Using regular expressions and file naming conventions allows reducing the number of required side car files. The same side car file could be used for all renditions for example. This allows the origin to reduce the number of sidecar files, but the edge will always receive several copies of the same data as caching is done on the exact filename. It is recommended to balance the advantages and disadvantages of regular expressions, because of its Central Processing Unit (CPU) load on the origin. The following is an example for a set of segments where the filenames satisfy the segmentRegex expression. In this example, the filenames are in the form of video_segment_[repID]_123.mp4, video_segment_[repID]_124.mp4 and so on, allowing to have one sidecar file for all Representations (for DASH). sidecar ( /version/ 1, /segments/ [{/segmentRegex/ "video_segment_ .*?_123.mp4", /position/ 21}, {/segmentRegex/ "video_segment_ .*?_124.mp4", /position/ 22}] ) When segments are byteranges: - sidecar shall contain only sidecar-byterange elements. - version is set to 1 for sidecar files compliant to the present document. - fileSize is the size of the track in bytes. - startRange defines the position of the first byte in the byterange. This expressed as a byte offset from the beginning of the track sidecar-byterange elements in the array shall be ordered in increasing startRange values. - position is defined in clause 5.5.2. NOTE 2: The first byterange of a track contains the initialization segment. When segments are delivered with byteranges, it is not possible to differentiate the request for this part of the file from a request for a media segment when using a pattern as described in clause 5.3. The initialization segment is not watermarked, therefore position equal -1 for this segment. The following is an example of a file with an initialization segment part of the byterange from 0 to 1117 and two segments. Sidecar ( /version/ 1, /fileSize/ 262445216, /segments/ [{/startRange/ 0, /position/ -1}, {/startRange/ 1118, /position/ 0}, {/startRange/ 1701212, /position/ 1}, β¦ {/startRange/ 261083393, /position/ 118}, {/startRange/ 262073936, /position/ 119}] ) |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.3 HTTP Header | When content is pushed, in the request header, under the WMPaceInfoIngest HTTP header field, the following JSON object is added: WMPaceInfoIngest : { "version": version, "variant": variant, "position": position, "firstpart": firstpart, "lastpart": lastpart } Where - version is set to 1 for WMPaceInfoIngest compliant to the present document. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 17 - variant, position, firstpart and lastpart are defined in clause 5.5.2. When content is pulled, in the response header, under the WMPaceInfoEgress HTTP header field, the following CBOR object, base64url-encoded as described in clause 5 of IETF RFC 4648 [7], is added: WMPaceInfoEgress : <sidecar-discrete> Where - sidecar-discrete is defined in clause 5.5.3.2 and contains exactly one discrete-segment object with data for that segment. Below is an example of the JSON element added in a WMPaceInfoIngest header field where the payload of the HTTP request contains the full segment of Variant A. { "version": 1, "variant": 0, "position": 33, "firstpart": true, "lastpart": true } |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.4 ISOBMFF Box | The format of WMPaceInfo class shall be: class WMPaceInfo { unsigned int(8) version; unsigned int(8) variant; unsigned int(1) emulation_1; unsigned int(15) position; unsigned int(1) emulation_2; unsigned int(1) firstpart; unsigned int(1) lastpart; unsigned int(5) reserved; } Where - version is set to 1 for WMPaceInfo compliant to the present document. - variant, position, firstpart and lastpart are defined in clause 5.5.2. - emulation_1, and emulation_2 are set to 1. Within an ISOBMFF file, the WMPaceInfo class shall be carried in the following box: Box Type: 'wmpi' Container: Top level box Mandatory: No Quantity: Zero or one aligned(8) class WMPaceInfoBox extends Box('wmpi') { WMPaceInfo(); } This box should be inserted only at the beginning of a segment, after the styp box and before the moof box, in order to facilitate content manipulation when padding it (see clause 5.7.5.1). ETSI ETSI TS 104 002 V1.1.1 (2023-08) 18 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.5 SEI Message | SEI messages are inserted in the stream with a specific syntax depending on the codec. [8] provides the syntax for AVC, HEVC and AV1 video codecs in Annex B. In these messages: - The UUID shall be equal to 0xbec4f824-170d-47cf-a826-ce008083e355. - The watermarking metadata is the WMPaceInfo data with the format defined for the class WMPaceInfo() in clause 5.5.3.4. This message should be inserted for the first frame of a segment to facilitate content manipulation when padding it (see clause 5.7.5.1). |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.5.3.6 TS Adaptation Field | Following clause U of [2], the format of the private adaptation field descriptor carrying the WMPaceInfo data is defined in Table 4. Table 4: WMPaceInfo descriptor Syntax No. of bits Mnemonic temi_WMPaceInfo_descriptor { af_descr_tag af_descr_length WMPaceInfo() } 8 8 40 uimsbf uimsbf uimsbf Where - af_descr_tag is an 8-bit field that identifies this AF descriptor. It is equal to 0xDF. - af_descr_length is an 8-bit field specifying the number of bytes of the AF descriptor immediately following af_descr_length field. - WMPaceInfo() is a 40-bit field that carries the information defined for the class WMPaceInfo() in clause 5.5.3.4. This message should be inserted for the first frame of a segment to facilitate content manipulation when padding it (see clause 5.7.5.1). |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6 Content Preparation | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.1 Introduction | Content preparation means the generation of A/B Variants of the segments followed by the push of content on the origin. It is under a workflow manager responsibility in case of VOD and fully automated for Live content. The encoder generates the different Variants of the adaptive content. The encrypted segments, the DASH manifest and HLS playlists are generated by the packager and pushed to the origin. A simplified flow is shown in Figure 3 for the case of Live content if the DASH-IF ingest protocol is used [i.1] (note that content protection steps are omitted for clarity). For encrypted content, Variants of every segment part of the same Representation may be encrypted using the same encryption method and with the same content key, meaning the same DRM license allows decrypting the A and B Variants. In addition to the Variants, the encoder also pushes WMPaceInfo that contain information allowing the packager and the origin to properly associate the pieces of Variants that are pushed to a bit position on the WM pattern. In such flow, the packager can aggregate multiple ingest segments into one egress segment, with the limitation that only ingest segments carrying the same position value can be aggregated together. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 19 Figure 3: Example of Live DASH content preparation workflow using the DASH-IF ingest protocol |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.2 Encoding Recommendations | This clause contains recommendation when encoding content. The goal is to facilitate the creation and management of A and B Variants in the delivery chain. When segments are requested as byteranges in a file or when chunks are requested as byteranges in a segment, the segments and chunks in A and B Variants shall have the same size as the player receives only one DASH manifest or HLS playlist and will get byterange lengths from one sidx box only. How this is achieved in out of the scope of the present document (as an example, bit stuffing in the encoder is an option). NOTE 1: This solution does not allow creating aligned segment when content is delivered with HLS in the form of MPEG-2 TS segments encrypted with AES sample encryption, because start code emulation prevention is re-applied over the entire NAL unit after encryption with MPEG-2 TS. NOTE 2: An alternative solution is either to not use segments requested as byteranges, but to use discrete files (in these cases, there is no need to align Variant A and B of the same segment) or use CMAF segments with HLS where start code emulation prevention is not re-applied after encryption. 5.6.3 Delivering Content and WMPaceInfo from the Encoder to the Packager Only one option for conveying WMPaceInfo information from the encoder to the origin shall be used. Multiple concurrent formats are not allowed. NOTE 1: When WMPaceInfo is delivered in TS adaptation field, ISOBMFF box, or SEI, it adds overhead in the delivery from the CDN to devices. The sidecar file and HTTP header delivery methods do not. The encoder is sending part of segments to the packager, as the output of the encoder is not necessarily aligned on the segment length. Furthermore, when multiple streaming formats are used, it may happen that segments generated by the packager are not of the same size for every streaming protocol (for example, 2 seconds segments for DASH and 4 seconds segments for HLS). The encoder then needs a mechanism for announcing which parts of the Variants it sends can be aggregated in segments. This is achieved by using the firstpart and lastpart within WMPaceInfo. NOTE 2: Where an encoder delivers additional metadata to instruct the packager how to aggregate the content into segments, the encoder ensures that metadata and firstpart and lastpart fields are consistent. For example, the encoder could output the series of content elements of 1 second length with WMPaceInfo as shown in Figure 4. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 20 Figure 4: Example of output of an encoder If the encoder pushes over HTTP these elements, each one should carry a WMPaceInfoIngest HTTP header with the relevant data. Every server keeps the information within the header associated to the ingested segment. In some cases, for example when the origin does additional packaging, the header may be updated. The packager can then prepare segments according to the streaming protocol. From the example above, it can create segments of 2 or 4 seconds keeping the consistency of the watermarking. NOTE 3: In this case, 2 consecutive segments of 2 seconds carry the same position value, hence a larger piece of content is required to retrieve an identifier compared to the case where 2 consecutives segments carrying different position values. Other options are to carry WMPaceInfo in a sidecar file or SEI or ISOBMFF box or TS adaptation field. For cases where the origin can perform additional manipulation of the content, WMPaceInfo may be carried within the content instead providing it is overwritten as specified in clause 5.6.5. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.4 Segment Ingress Path Structure on the Origin | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.4.1 Introduction | The DASH manifest [1] and HLS playlist [3] served to the devices are "neutral", meaning that: - The same playlist or manifest is served to all devices of all end-users. - It does not expose different names for A and B Variants of a given segment. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.4.2 Locating the Variants | Egress DASH manifests and HLS playlists shall be neutral, but ingest DASH manifests and HLS playlists include information about the A and B Variants being ingested, this is: - The ingest path. - Some signalling elements to describe if a DASH Adaptation Set includes the A or B Variants, or if an HLS media playlist includes A or B Variants. The ingest of A and B Variants shall use specific ingest paths that include a Variant identification (${variantId}). DASH Ingest manifests shall include an AdaptationSet per Variant. The contents of the AdaptationSet shall be identical for every Variant apart from an EssentialProperty element that indicates the variantId and that the Variants are grouped (i.e. they reference the same media). It has the @schemeIdUri attribute equal to http://dashif.org/guidelines/watermarking_variant#${variantId} where ${variantId} identifies the Variant with which this EssentialProperty element is associated and @value attribute identifies the group to which the Variant belongs. If there are additional Variants (A, B and C for example), the @schemeIdUri attribute is different for each Variant, for example, for Variant C, @schemeIdUri attribute shall be equal to http://dashif.org/guidelines/watermarking_variant#c, if the schema with lower case letters is used. The following is an example of a DASH ingest manifest with two Variants, A and B. The watermarking signalling is highlighted in bold. EssentialProperty elements indicate that Variant A and Variant B belong to the same group ("tv1"). In this example, lower case letters are used for variantId. NOTE 1: Segment file naming with template based on segment $number or $time are possible. <AdaptationSet mimeType="video/mp4" segmentAlignment="true" startWithSAP="1" subsegmentAlignment="true" subsegmentStartsWithSAP="1" bitstreamSwitching="true"> 1 second firstpart:0 lastpart:1 position:2 firstpart:1 lastpart:0 position:3 firstpart:0 lastpart:0 position:3 firstpart:0 lastpart:0 position:3 firstpart:0 lastpart:1 position:3 firstpart:1 lastpart:0 position:4 firstpart:0 lastpart:0 position:4 4 seconds ETSI ETSI TS 104 002 V1.1.1 (2023-08) 21 <EssentialProperty schemeIdUri="http://dashif.org/guidelines/watermarking_variant#a" value="tv1"/> <SegmentTemplate timescale="60000" media="a/video_segment_$RepresentationID$_$Time$.mp4" initialization="a/video_init_$RepresentationID$.mp4" startNumber="10967120" presentationTimeOffset="903486496960"> <SegmentTimeline> <S t="903487696960" d="240000"/> <S t="903487936960" d="186000"/> </SegmentTimeline> </SegmentTemplate> <Representation id="27" width="1920" height="1080" frameRate="30/1" bandwidth="5000000" codecs="avc1.4D4028"/> <Representation id="24" width="1280" height="720" frameRate="30/1" bandwidth="3000000" codecs="avc1.4D401F"/> <Representation id="26" width="640" height="360" frameRate="30/1" bandwidth="1499968" codecs="avc1.4D401E"/> </AdaptationSet> <AdaptationSet mimeType="video/mp4" segmentAlignment="true" startWithSAP="1" subsegmentAlignment="true" subsegmentStartsWithSAP="1" bitstreamSwitching="true"> <EssentialProperty schemeIdUri="http://dashif.org/guidelines/watermarking_variant#b" value="tv1"/> <SegmentTemplate timescale="60000" media="b/video_segment_$RepresentationID$_$Time$.mp4" initialization="b/video_init_$RepresentationID$.mp4" startNumber="10967120" presentationTimeOffset="903486496960"> <SegmentTimeline> <S t="903487696960" d="240000"/> <S t="903487936960" d="186000"/> </SegmentTimeline> </SegmentTemplate> <Representation id="27" width="1920" height="1080" frameRate="30/1" bandwidth="5000000" codecs="avc1.4D4028"/> <Representation id="24" width="1280" height="720" frameRate="30/1" bandwidth="3000000" codecs="avc1.4D401F"/> <Representation id="26" width="640" height="360" frameRate="30/1" bandwidth="1499968" codecs="avc1.4D401E"/> </AdaptationSet> For HLS ingest playlists, the multivariant playlist shall include all the A and B Variants with a custom attribute specifying the Variant (using ${variantId} identification as defined in clause 5.3). The attribute is WATERMARKING- VARIANT. A combination of both audio and video watermarking can therefore be used in a single streamset. In the media playlists, the only specific signalling is the segments paths that reflects on which ingest path the Variants are ingested. The sub-paths in the media playlists shall use the same convention that the ${variantId}. The following is an example of HLS ingest playlists, the watermarking signalling is highlighted in bold (this theoretical example, both the video and audio are watermarked). In this example, lower case letters are used for variantId. Multivariant playlist #EXTM3U #EXT-X-VERSION:4 #EXT-X-INDEPENDENT-SEGMENTS #EXT-X-STREAM-INF:BANDWIDTH=5227200,AVERAGE- BANDWIDTH=3511200,CODECS="avc1.4d401f,mp4a.40.2",RESOLUTION=1280x720,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="a" video_1.m3u8 #EXT-X-STREAM-INF:BANDWIDTH=2719200,AVERAGE- BANDWIDTH=1861200,CODECS="avc1.77.30,mp4a.40.2",RESOLUTION=640x360,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="a" video_2.m3u8 #EXT-X-STREAM-INF:BANDWIDTH=8571200,AVERAGE- BANDWIDTH=5711200,CODECS="avc1.4d4028,mp4a.40.2",RESOLUTION=1920x1080,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="a" video_3.m3u8 #EXT-X-STREAM-INF:BANDWIDTH=5227200,AVERAGE- BANDWIDTH=3511200,CODECS="avc1.4d401f,mp4a.40.2",RESOLUTION=1280x720,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="b" video_4.m3u8 ETSI ETSI TS 104 002 V1.1.1 (2023-08) 22 #EXT-X-STREAM-INF:BANDWIDTH=2719200,AVERAGE- BANDWIDTH=1861200,CODECS="avc1.77.30,mp4a.40.2",RESOLUTION=640x360,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="b" video_5.m3u8 #EXT-X-STREAM-INF:BANDWIDTH=8571200,AVERAGE- BANDWIDTH=5711200,CODECS="avc1.4d4028,mp4a.40.2",RESOLUTION=1920x1080,FRAME- RATE=30.000,AUDIO="program_audio",WATERMARKING-VARIANT="b" video_6.m3u8 #EXT-X-IMAGE-STREAM-INF:BANDWIDTH=55649,AVERAGE- BANDWIDTH=23579,RESOLUTION=308x174,CODECS="jpeg",URI="trickplay_7.m3u8" #EXT-X-MEDIA:TYPE=AUDIO,LANGUAGE="eng",NAME="Stadium ambiance",AUTOSELECT=YES,DEFAULT=YES,GROUP- ID="program_audio",URI="audio_8.m3u8",WATERMARKING-VARIANT="a" #EXT-X-MEDIA:TYPE=AUDIO,LANGUAGE="eng",NAME="Stadium ambiance",AUTOSELECT=YES,DEFAULT=YES,GROUP- ID="program_audio",URI="audio_9.m3u8",WATERMARKING-VARIANT="b" NOTE 2: While it is a legal signalling in HLS to have multiple EXT-X-MEDIA tags with the same GROUP_ID value, each tag has a different NAME value. As these playlists are not for devices to consume and to minimize the processing on the playlists, the ingest playlists do not follow this rule and multiple EXT-X-MEDIA share the same NAME value. Media playlist (A Variant) #EXTM3U #EXT-X-VERSION:6 #EXT-X-INDEPENDENT-SEGMENTS #EXT-X-TARGETDURATION:6 #EXT-X-MEDIA-SEQUENCE:11352692 #EXT-X-MAP:URI="video_init_1.mp4" #EXT-X-PROGRAM-DATE-TIME:2021-09-15T00:48:38.933Z #EXTINF:6.000, a/video_segment_1_11352692.mp4 #EXTINF:6.000, a/video_segment_1_11352693.mp4 #EXTINF:6.000, a/video_segment_1_11352694.mp4 #EXTINF:6.000, a/video_segment_1_11352695.mp4 #EXTINF:6.000, a/video_segment_1_11352696.mp4 Media playlist (B Variant) #EXTM3U #EXT-X-VERSION:6 #EXT-X-INDEPENDENT-SEGMENTS #EXT-X-TARGETDURATION:6 #EXT-X-MEDIA-SEQUENCE:11352692 #EXT-X-MAP:URI="video_init_1.mp4" #EXT-X-PROGRAM-DATE-TIME:2021-09-15T00:48:38.933Z #EXTINF:6.000, b/video_segment_1_11352692.mp4 #EXTINF:6.000, b/video_segment_1_11352693.mp4 #EXTINF:6.000, b/video_segment_1_11352694.mp4 #EXTINF:6.000, b/video_segment_1_11352695.mp4 #EXTINF:6.000, b/video_segment_1_11352696.mp4 When the ingested content is not watermarked anymore, then: - For DASH content, the EssentialProperty elements shall be removed from the ingest manifest and a new Period shall be created with a single AdaptationSet. The path to the segments shall be updated, removing any information on the Variant location (in the example above, the a/ shall be removed from the @media value of the SegmentTemplate element). ETSI ETSI TS 104 002 V1.1.1 (2023-08) 23 - For HLS content, the encoder shall create a new multivariant playlist that does not include WATERMARKING- VARIANT attributes. It also stops delivering the additional media playlists for the B Variant and others if present. The path to the segments in the media playlist delivered to devices shall be updated, removing any information on the Variant location (in the example above, the a/ shall be removed from the media playlist). NOTE 3: Stopping watermarking content is different from toggling edge sequencing logic (see clause 5.3). |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.4.3 Locating the Sidecar File | The sidecar file is part of the ingest with the DASH manifest or HLS playlist, the link to this file is added in different places depending on the format. DASH ingest manifests shall include an EssentialProperty element at the Representation level with a @schemeIdUri attribute equal to http://dashif.org/guidelines/watermarking_wmpaceinfo and @value attribute equal to the pointer to the sidecar file. The pointer is relative to the ingest manifest. The following is an example of a DASH ingest manifest where the watermarking signalling is highlighted in bold. In this example, the absolute path for the sidecar file for the first representation is equal to https://dash.edgesuite.net/dash264/TestCases/1a/ElephantsDream_H264BPL30_0100.264.dash_wm_pace_info. NOTE: This example also includes the signalling defined in clause 5.6.2 (for one Variant A). In this case, the EssentialProperty elements are added in the Representation. <?xml version="1.0" encoding="UTF-8"?> <MPD xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="urn:mpeg:dash:schema:mpd:2011" xsi:schemaLocation="urn:mpeg:dash:schema:mpd:2011 DASH-MPD.xsd" type="static" mediaPresentationDuration="PT654S" minBufferTime="PT4S" β¦ <AdaptationSet mimeType="video/mp4" codecs="avc1.42401E" subsegmentAlignment="true" subsegmentStartsWithSAP="1" contentType='video' maxWidth="480" maxHeight="360" maxFrameRate="24" par="4:3"> <Representation id="2" bandwidth="150000" width="480" height="360" frameRate="24"> <EssentialProperty schemeIdUri="http://dashif.org/guidelines/watermarking_variant#a" value="tv1"/> <EssentialProperty schemeIdUri="http://dashif.org/guidelines/watermarking_wmpaceinfo" value="ElephantsDream_H264BPL30_0100.264.dash_wm_pace_info"/> <BaseURL>a/ElephantsDream_H264BPL30_0100.264.dash</BaseURL> <SegmentBase indexRange="984-11244"> <Initialization range="0-983"/> </SegmentBase> </Representation> <Representation id="3" bandwidth="250000" width="480" height="360" frameRate="24"> <EssentialProperty schemeIdUri="http://dashif.org/guidelines/watermarking_variant#a" value="tv1"/> <EssentialProperty schemeIdUri="https://dashif.org/guidelines/watermarking_wmpaceinfo" value="ElephantsDream_H264BPL30_0175.264.dash_wm_pace_info"/> <BaseURL>a/ElephantsDream_H264BPL30_0175.264.dash</BaseURL> <SegmentBase indexRange="984-11245"> <Initialization range="0-983"/> </SegmentBase> </Representation> β¦ </AdaptationSet> </MPD> HLS ingest playlists shall include in the media playlist a custom tag specifying the pointer to the sidecar file. The pointer is relative to the ingest manifest. The tag is #EXT-X-WMPACEINFO:<attribute-list> where the defined attribute is URI, a quoted-string that gives the relative pointer to the sidecar file. In the media playlist for each Variant (A, B, C β¦), the sidecar file referenced by the #EXT-X-WMPACEINFO tag is the same as the variant value shall not be considered. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 24 The following is an example of a HLS media playlist, the watermarking signalling is highlighted in bold. Note that the multivariant playlist remains unmodified. #EXTM3U #EXT-X-TARGETDURATION:8 #EXT-X-VERSION:7 #EXT-X-MEDIA-SEQUENCE:1 #EXT-X-PLAYLIST-TYPE:VOD #EXT-X-INDEPENDENT-SEGMENTS #EXT-X-WMPACEINFO:URI="main_wm_pace_info" #EXT-X-MAP:URI="main.mp4",BYTERANGE="1118@0" #EXTINF:7.98333, #EXT-X-BYTERANGE:1700094@1118 a/main.mp4 #EXTINF:8.00000, #EXT-X-BYTERANGE:1789481@1701212 a/main.mp4 #EXTINF:8.00000, #EXT-X-BYTERANGE:1777588@3490693 a/main.mp4 #EXTINF:8.00000, #EXT-X-BYTERANGE:1752144@5268281 a/main.mp4 #EXTINF:7.26667, #EXT-X-BYTERANGE:1563219@7020425 a/main.mp4 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.6.5 Packaging Recommendations | This clause contains requirements where packaged content is served to devices. The goal is to facilitate the creation and management of A and B Variants in the delivery chain. These requirements apply even if no re-packaging process exists. NOTE: This implies that an encoder working against a completely passive receiver (e.g. interface 2 of [i.1]) publishes egress versions of the content directly. The minimum segment duration should consider the embedding capabilities of the WM technology in order to ensure that a segment contains only information for A or B Variant. A segment carrying only one bit of information (Variant A or B) allows to match a segment to a bit value in the WM pattern. As described in clause 5.6.3, a re-packaging process may aggregate received parts of content. It builds a segment beginning with the part of content with firstpart=true and then aggregates until lastpart=true for creating a segment until the targeted length has been reached. It shall begin creating a new segment if a part of content with firstpart=true is received before reaching the targeted length. The packager shall not aggregate segments that have inconsistent metadata, more precisely, only ingest segments carrying the same position value shall be aggregated together. The transformation of ingest manifest into egress manifests requires the following actions: - All watermarking_wmpaceinfo and watermarking_variant EssentialProperty elements in DASH manifests and EXT-X-WMPACEINFO tags in HLS playlists shall be removed from the egress manifests. - HLS media playlists of a given rendition in HLS shall be merged into a single, neutral version of it (without ${variantPath}). - DASH manifests shall be made neutral (without ${variantPath}). While the manifests are made neutral when delivered to devices, the content shall remain stored with the structure defined in the ingest manifests. Doing so, when the CDN edge requests a Variant for a given segment in applying the logic defined in clause 5.7.5, the origin has a direct access to the requested Variant. Additionally, when translating from ingress to egress, a re-packaging process shall: - overwrite WMPaceInfo when carried as SEI messages, TS adaptation fields or ISOBMFF boxes. Overwriting shall prevent start code emulation. It is recommended to overwrite with 0xFF; ETSI ETSI TS 104 002 V1.1.1 (2023-08) 25 - remove firstpart, lastpart, segmentRegex from sidecar-discrete elements. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7 Content Playback | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.1 Introduction | The flow for content playback is shown in the following clauses. The origin received content as explained in clause 5.5. It has access to the A/B Variants and the WMPaceInfo data. This clause describes only the case where the WM token is used in direct mode and does not consider the value of wmsegduration (hence using WMPaceInfo). This clause is also not considering the case of download of content for later offline playback. Usually, content available for download is available in the form of byteranges and the device requests large byteranges that overlap those announced in the MPD or HLS playlists. When content is watermarked, this is not possible as only announced byteranges are addressable (see clause 5.7.5.4). The device shall therefore either use the announced byteranges only or a proxy shall ensure that the edge receives requests that are for announced byteranges. Content playback is divided in three actions: - Acquiring the WM token, the DASH manifest, or the HLS playlists - Acquiring the initialization segment - Acquiring media segments While the first action is common to all type of content, the other ones have variations depending on the packaging and delivery mode of the content. Variation is, for example on the difference between content delivered as byterange or discrete segments. Another possible variation appears when HLS low latency is used for the chunks requested at the edge of live. The following goes through the different actions by providing the expected workflows. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.2 Dynamic Ad Insertion | In case of Dynamic Ad Insertion (DAI), the break may happen at any time. As every segment carries watermarking information allowing to perform the detection, there shall not be segments carrying conflicting data. While some techniques may recover from this mix of data, it will, in all cases, impact the length of content needed for retrieving the unique identifier. For Live content, assuming that an ad replacement period is defined, then from the device perspective, the following consumption modes are possible. - The device consumes ads from an alternative edge for the full duration of the ad break - The device consumes ads from an alternative edge for a duration shorter than the replacement period - The device consumes the original content as no replacement ad is proposed Devices may therefore consume content differently during the ad break. For VOD content, ads will be inserted or stitched with ad break (cue in/out points for example) markers. The device should consume them from an alternative edge for the full duration of the ad break. The encoder shall watermark ads part of the original content for Live content. The watermarking technology shall remain consistent between all these options. Some devices may receive the original content if no ad can be found for replacement. One consequence is that these devices receive content that is meant to be watermarked following the rules of the present document. Devices receiving an ad for replacement shall receive it from a different edge that does not enforce watermarking. Such edge will then gracefully ignore the WM token. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 26 The WM token is expected to be present in all playback requests during the session. In presence of a DAI manifest manipulator, depending on its behaviour, it may be necessary to tweak the configuration of the delivery pipeline to guarantee the propagation of the WM token. For instance, it may be required to perform some manifest manipulation at the edge to re-introduce the WM token in the response, e.g. when the token is transported as a query parameter and the DAI manifest manipulator is not piggybacking incoming query parameters in the rewritten manifest/playlist. Another case is when the watermark token is incorporated to the virtual path, stripped at the edge on its way to the DAI manifest manipulator (that remains therefore unaware of the WM token) which returns a manipulated playlist that contains absolute URLs. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.3 WM Token, DASH Manifest and HLS Playlists Acquisition | The device acquires the WM token in an implementation specific manner. It may be retrieved directly from a WM token server, or it may be provided in a response from another server as part of other data required for playing back content. The WM token may be added as part of the virtual path of the requested object, as a query string attribute or as part of the HTTP header when the device requests content to the edge. It is recommended to use the virtual path. The WM token may be added by the device for requesting DASH manifest and HLS playlists. While these objects are not watermarked (the pattern in the name allows the edge to know this), the edge may validate or not the token and refuse to serve these objects if the token is not valid. The edge may also gracefully ignore the token. The origin cleans the served objects, removing any property related to location of objects (see clause 5.6.5). The manifest and playlist are neutral. This is summarized in Figure 5. Figure 5: Token, DASH manifest and HLS playlist acquisition ETSI ETSI TS 104 002 V1.1.1 (2023-08) 27 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.4 Initialization Segment Acquisition | When content is delivered as byteranges, as the initialization segment is within the file, the token shall be added in the request as the requested file has a name that matches the pattern for watermarked content. The edge will then apply the exact same logic it applies for a media segment, it retrieves the sidecar file and extracts the WMPaceInfo for the first part of the track that contains the initialization segment (as defined in clause 5.7.5). It can then deliver the initialization segment to the device. As position is equal to -1 (not watermarked), it shall deliver the initialization segment from Variant A. One or several Variants may become unavailable on the origin for any reason, such as a lost connection with the encoder for these encoding pipelines. Such situation will result in a failed playback if Variant A is the one that is not available. The origin shall deliver to the edge the initialization segment from any available Variant in this case on the endpoint for Variant A. NOTE: The token is evaluated and validated as the edge cannot make a difference between the initialization segment and a media segment. When content is delivered as discrete segments, the name of the initialization segment shall not match the pattern for watermarked content as written in clause 5.3. The WM token may be added by the device for requesting the initialization segment. The edge may validate it or not and may refuse to serve these objects if it is not valid. The edge may also gracefully ignore it. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.5 Media Segments and WMPaceInfo Acquisition | |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.5.1 General Requirements | For the media segments, a token shall be attached to the HTTP requests. If not present, the edge shall reject the request and shall not deliver the segment. The edge shall validate the WM token (that can include checking signed data or decrypting some claims) which is attached to the requests and extracts the WM pattern so that the correct Variant can be sequenced. Watermarked objects shall include in the sub-path in the edge forward requests to the origin the value of identifying Variants that is part of the configuration described in clause 5.3. A request received at the CDN edge for https://edge.hostname/path/to/endpoint/video_segment_5_8353305.mp4 shall be translated into a forward request for https://origin.hostname/path/to/endpoint/${variantPath}video_segment_5_8353305.mp4 where the value of ${variantPath} depends on the value extracted from the WM pattern for this segment. The same logic applies if the watermarking is done through audio segments. The connection between the origin and the edge shall be restricted to legitimate requests. How this is achieved is out of the scope of the present document. NOTE 1: A static secret (a shared key), dynamic signatures or access lists (based on IP addresses) are examples of tools for restricting the access. There may be the need to disable watermarking within or upstream of the packager at any time, for example, one or several Variants may become unavailable on the origin for any reason, such as a lost connection with the encoder for these encoding pipelines. As devices request all Variants, this situation will result in intermittent black screens when requesting the affected Variants. In such case, position shall be set to -1 in WMPaceInfo, effectively announcing to the edge sequencing logic that segments are not watermarked. The edge shall then consume segment on the endpoint for Variant A. If this endpoint is not working properly, the origin shall deliver any available Variant on this endpoint. NOTE 2: This is breaking the watermarking detection. The period when such contingency measure is applied is not to be used for detection. How the end-to-end system is synchronized is out of the scope of the present document. As an example, the origin can raise an alarm. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 28 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.5.2 WMPaceInfo Acquisition | For each device request for /pathname/filename, the edge shall retrieve from the origin egress WMPaceInfo data associated to this object. The origin presents this information differently whether segments are discrete or byteranges: - For byterange segment, the origin shall have a dedicated endpoint for delivering WMPaceInfo information as a sidecar file. For a segment requested by a device at /pathname/filename, the origin shall have an endpoint /pathname/WMPaceInfo/filename that makes the sidecar file available. The response payload shall contain the sidecar file (as defined in clause 5.5.3.2 for byterange segments). The origin shall not extract data and only provide the sidecar file to the edge. The Content-Type for this object is application/cbor. - For discrete segment, the origin: - Shall have a dedicated endpoint /pathname/WMPaceInfo/filename for delivering WMPaceInfo for the requested segment. The response payload shall contain a sidecar file that contain a single WMPaceInfo object. The Content-Type for this object is application/cbor. - Shall add WMPaceInfo in the response header (as defined in clause 5.5.3.3) under the WMPaceInfoEgress header field when the edge requests the segment. - It is the edge that defines which endpoint it uses. If WMPaceInfo was delivered to the origin in ingress form (as part of the HTTP request headers, SEI message, ISOBMFF box, TS adaptation field or a sidecar file per track), that data shall be extracted and made available in egress form to the edge as both a HTTP header and dedicated endpoint. Any direct request from a device with /pathname/WMPaceInfo/filename shall receive an error code 403. Table 5 gives examples of content flows as ingest to the origin and egress of the origin to the edge. Table 5: Examples of content flows Live content VOD content Ingest of the origin No sidecar file, data is delivered as part of HTTP headers, SEI messages, ISOBMFF boxes or TS adaptation field. For both discrete segments and byteranges, one sidecar file per track. Egress of the origin One sidecar file per segment (note the special case of HLS low latency with byterange where multiple chunks are be linked to the same sidecar file, see clause 5.7.5.4) and HTTP header. For discrete segments, one sidecar file per segment and HTTP header. For byterange, one sidecar file per track. There are then three endpoints on the origin: - WMPaceInfo: /pathname/WMPaceInfo/filename - Variant A: /pathname/${variantPath}filename - Variant B: /pathname/${variantPath}filename Where ${variantPath} is as defined in clause 5.3. NOTE: Adding Variants creates additional endpoints. |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.5.3 Discrete Files | For the media segments delivered as discrete files, the flow is shown in Figure 6. The edge sequences the A or B Variant of a segment based on the WM pattern contained in the token. It has two options to know the position of the segment within the WM pattern: - First make a request to the origin to retrieve the WMPaceInfo data. This is done with a GET request using the path /pathname/WMPaceInfo/filename. The origin provides the WMPaceInfo from the Variant A in the payload of the response as a sidecar file. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 29 - Once, the data in WMPaceInfo is interpreted in conjunction with the WM pattern, the edge can request to the origin the right Variant corresponding to the position in the WM pattern that matches the value of position in WMPaceInfo and then deliver it to the device. - Make a request for the A and B Variants, extract the WMPaceInfo from one response header and once, the data in WMPaceInfo is interpreted in conjunction with the WM pattern, the edge can deliver the right Variant to the device. NOTE: There is a high probability that the edge will request both A and B Variants, hence adding WMPaceInfo to the response header allows avoiding an extra request to the origin. The edge caches the Variants of a given segment with different cache keys and it should prevent the cache keys to be revealed through debug headers. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 30 Figure 6: Media segment, as discrete file, acquisition ETSI ETSI TS 104 002 V1.1.1 (2023-08) 31 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.7.5.4 Byterange | For the media segments delivered as byteranges, the flow is shown in Figure 7. The edge delivers the A or B Variant of a segment based on the WM pattern contained in the token. To know which position in the WM pattern it has to consider, it needs to retrieve the sidecar file associated to this track. It first makes a HTTP GET request to the origin in order to retrieve the sidecar file. Whilst sub ranges within segments, such as chunks, are allowed, the edge shall not deliver byteranges overlapping several segments with different position values in WMPaceInfo. NOTE 1: An example is content delivered with HLS using the EXT-X-PART tag are byterange requests within a discrete segment. When the edge receives the request for this partial segment, it will request WMPaceInfo to the origin and will receive a sidecar file with only one WMPaceInfo. This allows the edge to know that it does not have enforce byterange validation for these requests. NOTE 2: Only byteranges overlapping valid ranges are problematic, requests for byteranges included in an allowed range are not breaking the WM pattern that is created by the A/B Variants and thus can be served. Once the data in WMPaceInfo is interpreted in conjunction with the WM pattern, the edge can deliver the correct Variant corresponding to the position in the WM pattern that matches the value of position in WMPaceInfo. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 32 Figure 7: Media segment, as byterange, acquisition ETSI ETSI TS 104 002 V1.1.1 (2023-08) 33 |
4b5f77d85580decea5022e899016dcb5 | 104 002 | 5.8 Monitoring and Watermark Detection | If content is found, a detection of a WM pattern can be performed. A video acquisition that includes valuable content (no commercial breaks for example) is performed. As the unique ID is obtained by extracting information from segments (0 or 1 in every segment), the acquired content shall be of several minutes (the longer the segments are, the longer the acquired video is). The video is then processed by the watermarking provider in order to extract the unique ID. This ID is then provided to the relevant entity that can match it to a device, user or streaming session and take the desired actions. How the detection is performed, and the revocation of the WM token is performed are out of the scope of the present document. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 34 Annex A (normative): Vendor Specific Core API A.1 Introduction In case of a token in indirect mode, it is expected that a vendor specific core (identified by wmvnd) generates the WM pattern (referred as wmpattern). This means that this requires some interaction between the edge and this vendor specific core. To facilitate this integration, the following defines the API made available by the vendor specific core. A.2 Edge-Vendor Specific API It is assumed that: - The call to the API function is blocking and the edge waits for the vendor specific core to end its processing. - The verification of the token is done before the call to the function. Verification includes the validation of the signature. The inputs are the values of the claims of the token that are relevant for the generation of the WM pattern. const crypto = require('crypto'); function generate_wmpattern (token.wmpatlen, token.wmkeyver, token.wmid, token.wmopid) { /* vendor specific processing */ return wmpattern; } ETSI ETSI TS 104 002 V1.1.1 (2023-08) 35 Annex B (informative): Examples of Workflows B.1 Introduction This annex takes the DASH-IF ingest protocol [i.1] as a reference. There are two interfaces defined: - Interface 1, where the combination of packager and origin is able to perform additional re-packaging hence the structure of ingest and egress may differ. Each POST/PUT contains one CMAF segment. This is often referred to an active receiving entity as a Just In Time Packager (JITP). - Interface 2, where the combination of packager and origin does not perform additional re-packaging, the structure of ingest and egress may be the same. The receiving entity is "passive", the source produces all objects in form that devices can consume. Each POST/PUT implicitly refers to one addressable object in an MPD or playlist. Therefore, the receiving entity is either active (interface 1) or passive (interface 2) and this leads to the following possibilities: - CMAF ingest, active receiving entity (JITP) - HLS/DASH ingest, active receiving entity (JITP) - HLS/DASH ingest, passive receiving entity Given all the options for carrying WMPaceInfo (see clause 5.5.3), the following describes some example flows for Live and VOD content. B.2 Live Content Flows For an active receiving entity (JITP), the grouping is non-trivial (as defined in [i.1] clause 6.2), therefore, as described in clause 5.6.4, the manifests are sent. The JITP may aggregate ingress segments according to (firstpart, lastpart) and WMPaceInfoEgress will reflect the aggregated result. In addition, evidence of WM process (such as the essential properties) is removed from egress playlists. If using the WMPaceInfoIngest header field on interface 1, the flow from the encoder to the edge is shown in Figure B.1. Figure B.1: Flow when using WMPaceInfoIngest and WMPaceInfoEgress header fields Another possible option is using sidecar file, this leads to the flow shown in Figure B.2. PUT/ingest-segment WMPaceInfoIngest: <json> Active Receiving Entity Edge Ingest Source Store ingest-segment & WMPaceInfoIngest GET/egress-segment WMPaceInfoEgress: <cbor> JITP translates WMPaceInfoIngest to WMPaceInfoEgress upon request from edge ETSI ETSI TS 104 002 V1.1.1 (2023-08) 36 Figure B.2: Flow when using WMPaceInfoIngest header field and sidecar file Another option is using SEI data. In this case, the receiving entity, either leaves WMPaceInfo in segment when storing and then overwrites it when serving after translating to WMPaceInfoEgress header or overwrites it before storing and saves the WMPaceInfo data somewhere else. The flow shown in Figure B.3. Figure B.3: Flow when using SEI data and WMPaceInfoEgress header field With a passive receiving entity, there is no media manipulation downstream of ingest source, therefore transferring WMPaceInfo data within the media is not an option, as it is not possible to overwrite it. Figure B.4 shows a possible flow with sidecar files. Figure B.4: Flow when using sidecar files PUT/ingest-segment WMPaceInfoIngest: <json> Active Receiving Entity Edge Ingest Source Store ingest-segment & WMPaceInfoIngest GET/WMPaceInfo/egress-segment GET/egress-segment JITP translates WMPaceInfoIngest to sidecar file upon request from edge PUT/ingest-segment SEI: WMPaceInfo() Active Receiving Entity Edge Ingest Source Store ingest-segment & SEI GET/egress-segment WMPaceInfoEgress: <cbor> JITP translates SEI to WMPaceInfoEgress upon request from edge PUT/egress-segment PUT/WMPaceInfo/egress-segment Passive Receiving Entity Edge Ingest Source Store egress-segment & Sidecar file GET/WMPaceInfo/egress-segment GET/egress-segment Encoder only sends egress WMPaceInfo in sidecar file when pushing to origin ETSI ETSI TS 104 002 V1.1.1 (2023-08) 37 B.3 VOD Content Flows If VOD content is prepared using live profile, then the permutations presented in clause B.2 are applicable. In addition, another option is that a single sidecar can describe all segments using regex for segmentRegex. This latter case leads to the flow shown in Figure B.5. Figure B.5: Flow when using sidecar files for VOD live profile If VOD content is prepared using on-demand profile, then the sidecar file is the only mechanism available to deliver WMPaceInfo data. This leads to the flow shown in Figure B.6. Figure B.6: Flow when using sidecar files for VOD on-demand profile For each segment of each representation PUT/segment After PUT/WMPaceInfo/sidecar PUT/manifest Receiving Entity Edge Ingest Source GET/egress-segment GET/WMPaceInfo/sidecar Store segments, sidecars and manifest For each representation PUT/trackfile PUT/WMPaceInfo/trackfile After PUT/manifest Receiving Entity Edge Ingest Source Store trackfiles, sidecars and manifest GET/manifest GET/WMPaceInfo/trackfile GET/trackfile Edge validates that byteranges do not overlap several segments with different position values ETSI ETSI TS 104 002 V1.1.1 (2023-08) 38 Annex C (normative): Registration Requests C.1 General This annex contains the registration requests for IANA (token claims) and MP4RA (4CC code). C.2 IANA Considerations The present document requests IANA to register the following claims in the following registry: https://www.iana.org/assignments/cwt/cwt.xhtml#claims-registry [12]. Version Claim - Claim Name: wmver - Claim Description: The version of the WM Token - JWT Claim Name: wmver - Claim Key: 300 - Claim Value Type: unsigned integer - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document Technology Vendor Claim - Claim Name: wmvnd - Claim Description: The WM technology vendor - JWT Claim Name: wmvnd - Claim Key: 301 - Claim Value Type: unsigned integer - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document Pattern Length Claim - Claim Name: wmpatlen - Claim Description: The length in bits of the WM pattern - JWT Claim Name: wmpatlen - Claim Key: 302 - Claim Value Type: unsigned integer - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document ETSI ETSI TS 104 002 V1.1.1 (2023-08) 39 Segment Duration Claim - Claim Name: wmsegduration - Claim Description: The nominal duration of a segment - JWT Claim Name: wmsegduration - Claim Key: 303 - Claim Value Type: map - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document Pattern Claim - Claim Name: wmpattern - Claim Description: The WM pattern - JWT Claim Name: wmpattern - Claim Key: 304 - Claim Value Types: COSE_Encrypt0 or COSE_Encrypt or byte string - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document ID Claim - Claim Name: wmid - Claim Description: Used as input to derive the WM pattern for indirect mode - JWT Claim Name: wmid - Claim Key: 305 - Claim Value Type: text string - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document Operator ID Claim - Claim Name: wmopid - Claim Description: Used as additional input to derive the WM pattern for indirect mode - JWT Claim Name: wmopid - Claim Key: 306 - Claim Value Type: unsigned integer - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document Key Version Claim - Claim Name: wmkeyver - Claim Description: The key to use for derivation of the WM pattern in indirect mode ETSI ETSI TS 104 002 V1.1.1 (2023-08) 40 - JWT Claim Name: wmkeyver - Claim Key: 307 - Claim Value Type: unsigned integer - Change Controller: DASH-IF - Specification Document(s): Clause 5.4 of the present document C.3 MP4RA Registration The present document requests MP4RA to register the following 4CC code. 1) The name, address, and URL of the organization requesting the code-point. DASH-IF 3855 SW 153rd Dr., Beaverton, OR 97003, USA https://dashif.org/ 2) The kind of code-point you wish to register (please choose from the set of registered types). Boxes (Atoms) 3) For all except object-type registrations, the suggested identifier (four-character code). Note that four- character codes use four 8-bit printable characters, usually from the first 128 Unicode characters (commonly thought of as plain ASCII), but at most from the first 256 Unicode characters. wmpi 4) The specification in which this code-point is defined, if possible. A copy of the specification would be appreciated, as it enables the authority to understand the registration better. If you are requesting a 'codec' code-point, a reference to the definition of the coding system itself, if separate from the definition of its storage in these files, would also be appreciated. Available from here (https://dashif.org/docs/IOP-Guidelines/DASH-IF-CTS-00XX-AB-Watermarking- 0.9.pdf). 5) A brief 'abstract' of the meaning of the code-point, perhaps ten to twenty words. wmpi stands for WaterMarkPaceInfo. It carries A/B forensic watermarking information within the ISOBMFF file. 6) Contact information for an authorized representative for the code-point, including: a) Contact person's name, title, and organization: DASH-IF Interoperability WG Chair b) Contact email: admin@dashif.org 7) Date of definition or implementation (if known) or intended date (if in future). July 31, 2023. 8) Statement of an intention to apply (implement) the assigned code-point. Expected to be implemented as part of DASH-IF conformance and reference tools according to the boilerplate in the specification. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 41 Annex D (informative): Code for Web Sequence Diagram D.1 Introduction This annex provides is the code for generating all workflows shown in Figures 3 to 7, excluding Figure 4 to be used on https://websequencediagrams.com [i.2]. D.2 Figure 3 Participant Encoder Participant Packager Participant Origin # STEP 1: Ingest from the encoder to the packager # For instance, the segmentation is 1s long Encoder -> Packager: Ingest manifest Encoder -> Packager: Ingest segments Variant A\n (w/ WMPaceInfo) Encoder -> Packager: Ingest segments Variant B\n (w/ WMPaceInfo) # STEP 2: Ingest from the Packager to the Origin (e.g. 2S long segments) # The Packager has to aggregate several DASH segments to produce the distributed segment Packager-> Origin: Egress manifest Packager-> Origin: Egress segments Variant A\n (w/ WMPaceInfo) Packager-> Origin: Egress segments Variant B\n (w/ WMPaceInfo) D.3 Figure 5 Participant Origin Participant CDN Edge Participant Device # STEP 1: Acquire a WM token opt WM token acquisition note over Origin,Device: Implementation specific end # STEP 2: Get the DASH manifest or HLS playlist for the viewing session alt Obtain DASH manifest Device->+CDN Edge: Get MPD(WM token) opt Manifest cache miss CDN Edge->+Origin: Get MPD Origin->Origin: Create a neutral MPD Origin-->-CDN Edge: MPD CDN Edge->CDN Edge: Cache MPD end CDN Edge-->-Device: MPD else Obtain HLS playlists Device->+CDN Edge: Get multivariant/media playlist(WM token) opt Multivariant/media playlist cache miss CDN Edge->+Origin: Get multivariant/media playlist Origin->Origin: Create neutral multivariant/media playlist Origin-->-CDN Edge: multivariant/media playlist CDN Edge->CDN Edge: Cache multivariant/media playlist end CDN Edge-->-Device: Multivariant/media playlist end ETSI ETSI TS 104 002 V1.1.1 (2023-08) 42 D.4 Figure 6 Participant Origin Participant CDN Edge Participant Device loop Segment request for playback Device->+CDN Edge: GET /pathname/segment_i(WM token) CDN Edge->CDN Edge: Validate WM token alt Invalid WM token CDN Edge-->Device: 401 Unauthorized else Valid WM token alt Use the dedicated endpoint for WMPaceInfo opt WMPaceInfo cache miss CDN Edge->+Origin: GET /pathname/WMPaceInfo/segment_i note right of Origin Origin retrieves WMPaceInfo for this segment and delivers it end note Origin-->-CDN Edge: 200 OK response CDN Edge ->> CDN Edge: Cache response end else Retreive WMPaceInfo from response header opt Variants cache miss CDN Edge->+Origin: GET /pathname/${variantPath}segment_i Origin-->-CDN Edge: 200 OK response CDN Edge ->> CDN Edge: Cache response CDN Edge->+Origin: GET /pathname/${variantPath}segment_i Origin-->-CDN Edge: 200 OK response CDN Edge ->> CDN Edge: Cache response end end alt Invalid Request: no WMPaceInfo for this segment CDN Edge-->Device: 400 Bad Request else Valid Request: WMPaceInfo available for this segment CDN Edge ->> CDN Edge: Create WMPaceInfoObject from cache CDN Edge ->> CDN Edge: VAR=getVariant(WM token, WMPaceInfoObject) alt If using the dedicated endpoint for WMPaceInfo opt Segment Variant cache miss CDN Edge->+Origin: GET /pathname/${VAR}/segment_i Origin-->-CDN Edge: 200 OK /pathname/${VAR}/segment_i CDN Edge ->> CDN Edge: Cache /pathname/${VAR}/segment_i end end CDN Edge-->Device: 200 OK with /pathname/segment_i(Variant ${VAR}) end end Device->Device: Play Content End D.5 Figure 7 Participant Origin Participant CDN Edge Participant Device loop Segment request for playback (including init segment) Device->+CDN Edge: GET /pathname/filename(WM token, byterange) CDN Edge->>CDN Edge: Validate WM token alt Invalid WM token CDN Edge-->Device: 401 Unauthorized else Valid WM token opt WMPaceInfo cache miss CDN Edge->+Origin: GET /pathname/WMPaceInfo/filename note right of Origin Origin retrieves WMPaceInfo sidecar file for this file and delivers it ETSI ETSI TS 104 002 V1.1.1 (2023-08) 43 end note Origin-->-CDN Edge: 200 OK response CDN Edge ->> CDN Edge: Cache response end alt Invalid Request: no WMPaceInfo for this file CDN Edge-->Device: 400 Bad Request else Valid Request: WMPaceInfo available for this file (one or many objects) CDN Edge ->> CDN Edge: Create WMPaceInfoObjects list from cache payload CDN Edge ->> CDN Edge: WMPaceInfoObject=getObject(WMPaceInfoObjects, byterange) alt Invalid byterange request CDN Edge-->Device: 400 Bad Request (Invalid byterange) else Valid byterange request CDN Edge ->> CDN Edge: VAR=getVariant(WM token, WMPaceInfoObject) opt Byterange cache miss CDN Edge->+Origin: Get /pathname/${VAR}/filename(byterange) note right of Origin The returned payload may be larger than the requested byterange (Partial Object Caching) end note Origin-->-CDN Edge: 206 Partial Content CDN Edge ->> CDN Edge: Cache /pathname/${VAR}/filename(byterange) end opt Partial Object Caching CDN Edge->>CDN Edge: Construct byterange response from locally cached object\n/pathname/${VAR}/filename(byterange) end CDN Edge-->Device: 206 Partial Content end end end Device->Device: Play Content End ETSI ETSI TS 104 002 V1.1.1 (2023-08) 44 Annex E (informative): Change History Date Version Information about changes 2022-03-23 0.8.0 Version published for first community review. 2023-02-02 0.9.0 Version published for second community review. 2023-02-09 0.9.1 Added IANA and MP4RA registration annexes. 2023-05-02 0.9.2 (Editorial) Corrections on broken automatic references and few formatting issues. (Editorial) Changed Master to Multivariant (HLS). (IANA) Updated the sidecar file integer keys and claim keys with final values. Removed remaining "must" from the text. Added COSE_Encrypt option for wmpattern. Updates on the variantPath construction options (removed the '.' possibility). Deleted examples for token claims. Changed the encryption algorithm for the pattern (align with CTA WAVE CAT). Clarified the storage paths for the Variant on the origin. Clarified the order of the bits in the WM pattern. 2023-05-03 0.9.3 Version published for IOP Review with some small editorial updates. 2023-05-09 0.9.5 Version created for IPR Review and ETSI Submission. ETSI ETSI TS 104 002 V1.1.1 (2023-08) 45 History Document history V1.1.1 August 2023 Publication |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 1 Scope | The present document defines O-RAN OAM interface functions and protocols for the O-RAN O1 interface. The present document studies the functions conveyed over the interface, including management functions, procedures, operations, and corresponding solutions, and identifies existing standards and industry work that can serve as a basis for O-RAN work. |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 2 References | |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found at https://docbox.etsi.org/Reference/. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. The following referenced documents are necessary for the application of the present document. [1] ETSI TS 128 314 (V17.0.0): "5G; Management and orchestration; Plug and Connect; Concepts and requirements (3GPP TS 28.314 version 17.0.0 Release 17)". [2] ETSI TS 128 315 (V17.0.0): "5G; Management and orchestration; Plug and Connect; Procedure flows (3GPP TS 28.315 version 17.0.0 Release 17)". [3] ETSI TS 128 532 (V17.3.0): "5G; Management and orchestration; Generic management services (3GPP TS 28.532 version 17.3.0 Release 17)". [4] ETSI TS 128 537 (V17.2.0): "5G; Management and orchestration; Management capabilities (3GPP TS 28.537 version 17.2.0 Release 17)". [5] ETSI TS 128 545 (V17.0.0): "5G; Management and orchestration; Fault Supervision (FS) (3GPP TS 28.545 version 17.0.0 Release 17)". [6] ETSI TS 128 550 (V17.1.0): "5G; Management and orchestration; Performance assurance (3GPP TS 28.550 version 17.1.0 Release 17)". [7] ETSI TS 128 622 (V17.1.1): "Universal Mobile Telecommunications System (UMTS); LTE; 5G; Telecommunication management; Generic Network Resource Model (NRM) Integration Reference Point (IRP); Information Service (IS) (3GPP TS 28.622 version 17.1.1 Release 17)". [8] ETSI TS 132 341 (V17.0.0): " Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; File Transfer (FT) Integration Reference Point (IRP); Requirements (3GPP TS 32.341 version 17.0.0 Release 17)". [9] ETSI TS 132 342 (V17.0.0): "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; File Transfer (FT) Integration Reference Point (IRP); Information Service (IS) (3GPP TS 32.342 version 17.0.0 Release 17)". [10] ETSI TS 132 404 (V17.0.0): "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; Performance Management (PM); Performance measurements; Definitions and template (3GPP TS 32.404 version 17.0.0 Release 17)". ETSI ETSI TS 104 043 V11.0.0 (2024-06) 11 [11] 3GPP TS 32.421 (V17.2.0): "Telecommunication management; Subscriber and equipment trace; Trace concepts and requirements". [12] 3GPP TS 32.422 (V17.4.0): "Telecommunication management; Subscriber and equipment trace; Trace control and configuration management". [13] 3GPP TS 32.423 (V17.2.0): "Telecommunication management; Subscriber and equipment trace; Trace data definition and management". [14] ETSI TS 132 432 (V17.0.0): "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; Performance measurement: File format definition (3GPP TS 32.432 version 17.0.0 Release 17)". [15] O-RAN TS O-RAN.WG1.OAD: "O-RAN Architecture Description". [16] O-RAN TS O-RAN.WG11.Security-Protocols-Specification: "O-RAN Security Protocols Specifications". [17] O-RAN TS O-RAN.WG11.Security-Requirements-Specifacation: "O-RAN Security Requirements and Controls Specification". [18] ONAP - Service: VES Event Listener Specification V7.2.1, January 16, 2021. [19] IETF RFC 6022 (October 2010): "YANG Module for NETCONF Monitoring". [20] IETF RFC 6241 (June 2011): "Network Configuration Protocol (NETCONF)". [21] IETF RFC 7950 (August 2016): "The YANG 1.1 Data Modeling Language". [22] IETF RFC 7951 (August 2016): "JSON Encoding of Data Modeled with YANG". [23] ETSI TS 128 623 (V17.2.2): " Universal Mobile Telecommunications System(UMTS); LTE; 5G; Telecommunication management; Generic Network Resource Model (NRM) Integration Reference Point (IRP); Solution Set (SS) definitions (3GPP TS 28.623 version 17.2.2 Release 17)". |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found at https://docbox.etsi.org/Reference. NOTE: While any hyperlinks included in this clause were valid at the time of publication, O-RAN cannot guarantee their long-term validity. The following referenced documents are not necessary for the application of the present document, but they assist the user with regard to a particular subject area. [i.1] 3GPP TR 21.905 (V17.0.0): Vocabulary for 3GPP Specifications". [i.2] ETSI TS 128 316 (V17.0.0): "5G; Management and orchestration; Plug and Connect; Data formats (3GPP TS 28.316 version 17.0.0 Release 17)". [i.3] 3GPP TS 28.531 (V17.1.0): "Management and orchestration; Provisioning". [i.4] ETSI TS 128 533 (V17.2.0): "5G; Management and orchestration; Architecture framework (3GPP TS 28.533 version 17.2.0 Release 17)". [i.5] 3GPP TS 28.552 (V17.4.0): "Management and orchestration; 5G performance measurements". [i.6] Void. ETSI ETSI TS 104 043 V11.0.0 (2024-06) 12 [i.7] ETSI TS 128 632 (V17.0.0): "Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; Inventory Management (IM) Network Resource Model (NRM) Integration Reference Point (IRP); Information Service (IS) (3GPP TS 28.632 version 17.0.0 Release 17)". [i.8] ETSI TS 132 346 (V17.0.0): "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; Telecommunication management; File Transfer (FT) Integration Reference Point (IRP): Solution Set (SS) definitions (3GPP TS 32.346 version 17.0.0 Release 17)". [i.9] ETSI TS 137 320 (V17.1.0): "Universal Mobile Telecommunications System(UMTS); LTE; 5G; Radio measurement collection for Minimization of Drive Tests (MDT); Overall description; Stage 2 (3GPP TS 37.320 version 17.1.0 Release 17)". [i.10] O-RAN TS O-RAN.WG3.O1-Interface-for-NearRT-RIC: "O1 Interface Specification for Near Real Time RAN Intelligent Controller". [i.11] O-RAN TS O-RAN.WG5.O-CU-O1.0: "O1 Interface Specification for O-CU-UP and O-CU-CP". [i.12] O-RAN TS O-RAN.WG5.WG5.O-DU-O1.0: "O1 Interface Specification for O-DU". [i.13] O-RAN TS O-RAN.WG10.WG10.OAM-Architecture: "O-RAN Operations and Maintenance Architecture". [i.14] O-RAN TS O-RAN.WG10.Information Model and Data Models.0: "O-RAN Information Model and Data Models Specification". [i.15] Void. [i.16] Void. [i.17] Void. [i.18] Recommendation ITU-T X.733: "Information technology - Open Systems Interconnection - Systems Management: Alarm reporting function". |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 3 Definition of terms, symbols and abbreviations | |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 3.1 Terms | For the purposes of the present document, the terms given in 3GPP TR 21.905 [i.1] apply. NOTE: A term defined in the present document takes precedence over the definition of the same term, if any, in 3GPP TR 21.905 [i.1]. |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 3.2 Symbols | For the purposes of the present document, the symbols given in 3GPP TR 21.905 [i.1] apply. NOTE: A symbol defined in the present document takes precedence over the definition of the same symbol, if any, in 3GPP TR 21.905 [i.1]. |
3c44d6b4e93165c24c893e9e439c2e16 | 104 043 | 3.3 Abbreviations | For the purposes of the present document, the abbreviations given in 3GPP TR 21.905 [i.1] and the following apply: NOTE: An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in 3GPP TR 21.905 [i.1]. 3GPP 3rd Generation Partnership Project ETSI ETSI TS 104 043 V11.0.0 (2024-06) 13 ASN.1 Abstract Syntax Notation One CM Configuration Management CRUD Create, Read, Update, Delete FS Fault Supervision FTPES File Transfer Protocol with Explicit SSL/TLS encryption GPB Google Protocol Buffers HTTP HyperText Transfer Protocol HTTPS HTTP Secure ID IDentifier IETF Internet Engineering Task Force IOC Information Object Class IP Internet Protocol JSON JavaScript Object Notation MDT Minimization of Drive Testing ME Managed Element MF Managed Function MnS Management Service MOC Managed Object Class MOI Managed Object Instance Near-RT RIC O-RAN Near Real Time RAN Intelligent Controller NETCONF NETwork CONFiguration protocol NF Network Function NGRAN Next Generation Radio Access Network NMS Network Management System NR New Radio NRM Network Resource Model O-CU-CP O-RAN Central Unit β Control Plane. O-CU-UP O-RAN Central Unit β User Plane O-DU O-RAN Distributed Unit O-RAN Open Radio Access Network O-RU O-RAN Radio Unit ONAP Open Network Automation Platform PM Performance Management or Performance Measurements PNF Physical Network Function RAN Radio Access Network RCEF RRC Connection Establishment Failure REST REpresentational State Transfer RFC Request For Comments RLF Radio Link Failure RRC Radio Resource Control SA5 Services & System Aspects Working Group 5 Telecom Management SBMA Services Based Management Architecture NOTE: See ETSI TS 128 533 [i.4], clause 4. SDO Standards Defining Organization SMO Service Management and Orchestration SFTP SSH File Transfer Protocol SSH Secure Shell TLS Transport Layer Security TR Technical Report TRS Trace Recording Session TS Technical Specification UE User Equipment URI Uniform Resource Identifier VES VNF Event Stream VNF Virtualized Network Function XML eXtensible Markup Language ETSI ETSI TS 104 043 V11.0.0 (2024-06) 14 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.