SLIDE 1
INTRODUCTION : Why certify?
- Video cameras / editors ubiquitous
- Data originates from multiple sources
- Consumer trusts some producers
- Must be able to validate provenance
- Examples:
Certifying Video Provenance Ashish Gehani I3P/SRI 1 INTRODUCTION - - PowerPoint PPT Presentation
Certifying Video Provenance Ashish Gehani I3P/SRI 1 INTRODUCTION - - PowerPoint PPT Presentation
Certifying Video Provenance Ashish Gehani I3P/SRI 1 INTRODUCTION : Why certify? Video cameras / editors ubiquitous Data originates from multiple sources Consumer trusts some producers Must be able to validate provenance
SLIDE 2
SLIDE 3
INTRODUCTION : Why embed?
- Auxiliary files:
– Consistent copies – Synchronized access – Storage overhead – Loose coupling → Error prone
- Embedding agnostic to:
– Administrative domain – Transfer between hosts, operating systems
- Video Embedding of Information for Lineage (VEIL):
– Inband encoding – Interoperable with legacy applications, libraries – Facilitate uptake
3
SLIDE 4
PROVENANCE : Desired properties
- Sound
– Signed elements
- Necessary
– Prune elements when possible
- Complete
– Don’t prune when output changes
4
SLIDE 5
PROVENANCE : Lineage tree
Output Operation Input 1 Input n
(a) Primitive operation (b) Compound operation tree
5
SLIDE 6
PROVENANCE : Granularity
- (Output, Executor, Input1, . . . , Inputn)
- Level - Assembler, System call, File?
- Fine → High overhead
- Coarse → False positives
- Can’t prune intra-process side-effects dynamically
6
SLIDE 7
PROVENANCE : Metadata format
- Primitive operation format:
IP Address Inode Time 1 Signature Output End Input n Executor Input
- Executor: 32 bit IPv4 address, 32 bit user ID
- Signature: 160 bits
[ S = SIGNKE(O, I1, . . . , In) ]
- Input / Output File:
– 32 bit IPv4 address – 32 bit inode – 32 bit time (Unix seconds count from 1 Jan, 1970)
7
SLIDE 8
PROVENANCE : Collection
- Manually:
> veil -o Output.mov -i Input 1.mov Input 2.mov Input 3.mov
- Automatically:
- pen()
File 1 Read File 2 Read close()
- pen()
- pen()
close() File 3 Write Process execution Time close()
8
SLIDE 9
RELATED WORK :
- Provenance
– Single host semantics
- Steganography
– Secrecy → Low capacity
- Watermarking
– Robustness → Low capacity
9
SLIDE 10
EMBEDDING : Interposition point
Signal to Frame Decomposition Motion Estimation VEIL Interpolation Residue Calculation Entropy Coding Decoding Entropy Reconstitution Frame Block VEIL Subpixel Analysis Decompression Interpolation Compression Quantization Spatial to Signal Domain Dequantization Spatial Domain 10
SLIDE 11
EMBEDDING : Overview
Metadata encoding operation
1 Signature Output End Input n Executor Input 1 Signature Output End Input n Executor Input 1 Signature Output End Input n Executor
- Primitive
Operation Video Frame
Input
11
SLIDE 12
EMBEDDING : Subpixel displacement
- Human visual system tolerates subpixel shift:
- riginal frame
- riginal frame
interpolated frame interpolated frame Object in Object in VEIL Outline of Outline of block in
12
SLIDE 13
EMBEDDING : Registration error compensation
Surface of object Points on surface mapped to first frame Points on surface mapped to second frame Points on surface mapped to third frame Camera Registration Grid
- v1, v2 : Adjacent pixel intensities
- vi : Interpolated pixel intensity
- vi = v1 + (δx − δi)(v2 − v1)
13
SLIDE 14
EXTRACTION : Checking lineage
Algorithm : CHECKLINEAGE(D) {E, S, O, I1, . . . , In} ← GETROOT(D) OUTPUT(E) PE ← PKILOOKUP(E) if I1, . . . , In = {} then VERIFYSIGNATURE(PE, S, O) else 8 > > > > > > > > < > > > > > > > > : Result ← VERIFYSIGNATURE(PE, S, O|I1| . . . |In) if Result = TRUE then 8 < : for i ← 1 to n do CHECKLINEAGE(Ii) else CheckFailed
14
SLIDE 15
EXTRACTION : Interpolation estimation
- Subpixel alignment minimizes visual distance:
dx
15
SLIDE 16
EXTRACTION : Calculating subpixel displacement
Theorem: δx can be calculated in O(β) time using O(log β) space, where β is block size. Proof: MSE(δx) = 1 β
xmax
- x=xmin
ymax
- y=ymin
[F(x + δx, y) − G(x, y)]2 d d(δx)MSE(δx) = 0 δx =xmax
x=xmin
ymax
y=ymin [ [F(x + 1, y) − F(x, y)] [F(x, y) − G(x, y)] ]
xmax
x=xmin
ymax
y=ymin [F(x + 1, y) − F(x, y)]2
- 16
SLIDE 17
CONCLUSION :
- – No auxiliary files to manage
– No storage overhead – Legacy decoders work
- – Fraction of data needed for lineage
– All data needed to bind lineage to data – Legacy writes detectable
- – Real-time speed
– 21 writers, 4 inputs each, 8x8 blocks, 4 bits/block – Tree extraction: 0.06 seconds
17
SLIDE 18
More?
18
SLIDE 19
DESIGN : Why not watermark?
- Security
– Watermark protects producer – VEIL protects consumer
- Robustness
– Watermark is: ∗ Robust - survives aggressive distortion ∗ Fragile - localizes distortion – Consumer not adversary, constraints irrelevant
- Visibility
– Watermark introduces flickering, frame noise – VEIL avoids visual distraction
19
SLIDE 20
DESIGN : Why not watermark?
[Continued]
- Capacity
– Video has heterogeneous signal, noise characteristics – Limited areas with: ∗ Low visual distortion ∗ High robustness ∗ High capacity – VEIL needs large capacity
20
SLIDE 21
PROVENANCE : Space requirement
- Space requirement (in KB):
Fan-in 1 2 3 4 Levels 2 0.09 0.14 0.19 0.24 3 0.14 0.34 0.65 1.05 4 0.19 0.75 2.02 4.30 5 0.24 1.56 6.13 17.30
21
SLIDE 22
EVALUATION : Extraction time
- δx extraction in C, Mac OS 10.4, 2 GHz Intel
- Time depends on block size:
Block size Time to compute δx (in µs) Overhead (in sec) for 1 sec of video (640x480 resolution, 30 fps) 4x4 1.3 0.723 8x8 5.2 0.746 16x16 18.2 0.648 32x32 72.2 0.649 64x64 292.1 0.657 22
SLIDE 23
EVALUATION : Embedding capacity
- Tree - Fan-in: 4, Levels: 4, Storage: 35, 264 bits
- I frames: 5%