Presentation Transcript
Temporal Synchronization in Video Watermarking Eugene T.Lin & Edward J.Dept: Temporal Synchronization in Video Watermarking Eugene T.Lin & Edward J.Dept Eugene T.Lin, Student Member, IEEE, and Edward J.Dept, Fellow, IEEE.
Outline: Outline Introduction
Watermark Embedding Model
Watermark Detection Model
Enhancement
Experiment
Introduction: Introduction
Introduction: Introduction Embed:
f0 f1 f2 f3 f4 f5 …
k0 k1 k2 k3 k4 k5 …
Detect:
… f …
… k? …
Introduction: Introduction Attack
Frame drop
Frame insert
Frame transposition
Frame averaging
Temporal interpolation or temporal scaling
Introduction: Introduction Key schedule
Time-invariant key
1 1 1 1 1 1 1 …
Time-periodic key
1 2 3 1 2 3 1 2 3 …
Time-independent key
1 2 3 4 5 6 7 8 9 …
Estimation attack?
Introduction: Introduction
Watermark Embedding Model: Watermark Embedding Model
Watermark Embedding Model: Watermark Embedding Model
Watermark Embedding Model: Watermark Embedding Model Key Generator
assume key space K = { 0, 1, 2 …, k-1 }
( S, S0, I, O, ψ, λ )
S : the set of states : { s0,s1,s2,s3 …, s|K|-1 }
S0 : initial state
I : input domain = K x S
O: output domain = K
ψ: transition function : S x I S
s(t+1) = ψ( s(t), Ke, F(t) )
λ: output function : S O
k(t) = λ( s(t) )
Watermark Embedding Model: Watermark Embedding Model
Watermark Detection Model: Watermark Detection Model
Watermark Detection Model: Watermark Detection Model
Watermark Detection Model: Watermark Detection Model Queue & preditor
ψ: transition function : S x I -> S
s(t+1) = ψ( s(t), Ke, F(t) )
λ: output function : S -> O
k(t) = λ( s(t) ) F0 F1 F4 F2 s0 s1 s2 s3 Queue
Watermark Detection Model: Watermark Detection Model Problem?
Frame insert ? OK!!
Frame drop ? Fail …
Watermark Detection Model: Watermark Detection Model Key schedule
111 222 333 1’1’1’ 2’2’2’ 3’3’3’
WEP
s(t+β) = ψ( s(t), Ke, F(t+β-1) )
Problem?
Enhancement: Enhancement
Enhancement: Enhancement WEP
s(t+β) = ψ( s(t), Ke, F(t+β-1) )
Problem ?
Enhancement: Enhancement Problem ?
… F1 F1 F1 F2 F2 F2 …
… k1 k1 k1 k1 k2 k2 …
Solve - Adaptive state transitions
Change key only
β freature vectors are the same , and
At lease β frames are embedded with same keys.
Enhancement: Enhancement Adaptive state transitions
Security
Define ψ as cryptographic hash function
S(t+1) = ψ(s(t),Ke,F(t)) = SH(s(t),Ke,F(t))
Nondeterministic SM
S(t+1) = ψ… = SH(s(t),Ke,F(t),R)
R 通常是很小的值。
Experiment: Experiment
Experiment: Experiment 3 key Schedule
Features only
Randomness only
Adaptive + Feature + Random
The End: The End