Signatures: Simulation and Disguise (An FHE Perspective)

(1)

Signatures: Simulation and Disguise (An FHE Perspective)

Muhammad Imran Malik

(2)

Signature Verification:

Forensic Handwriting Examiners' Perspective

Disturbed

Muhammad Imran Malik; Simulated and Disguised Signatures

18.12.2015 2

(3)

Physiological biometrics

18.12.2015 3

(4)

Behavioural biometrics

18.12.2015 4

(5)

But!

What do we know about the characteristics of natural writings?

And, in fact, handwriting in general

18.12.2015 5

(6)

Human movement

The properties of human movement form the theoretical underpinnings of forensic behavioural forensics

Evidence shows that there exists defined and well organised neural control in handwriting production

18.12.2015 6

(7)

Behavioural & physiological Order

Invariant features in handwriting:

A. Right (dominant) hand

B. Right arm with wrist immobilized C. Left hand

D. Pen gripped between the teeth E. Pen taped to the foot

18.12.2015 7

(8)

Behavioural & physiological Order

Vertical accelerations produced in writing the word ‘hell’.

One word has twice the amplitude as the other.

Note the temporal agreement but with difference in amplitudes of acceleration.

18.12.2015 8

(9)

Behavioural & physiological Order

Movement patterns are preserved when writing at different speeds

while keeping the writing size constant.

The dotted lines interpolating the times of occurrence of the major features, all have a common origin.

18.12.2015 9

(10)

Constancy of feature production

We observe constancy of feature production within a writer in spite of:

– Body position

– Hand movement within words

– Hand and arm movement between words – Hand and arm movement across lines – Hand and arm movement down the page

18.12.2015 10

(11)

Naturally written skilled signatures

Open Loop Control mechanism (OLC-mechanism) – Movement is performed in spite of afferent input

18.12.2015 11

(12)

Naturally written skilled signatures

– Examples

18.12.2015 12

(13)

Naturally written skilled signatures

– Further examples

18.12.2015 13

(14)

Naturally written skilled signatures

– An example naturally written signature

18.12.2015 14

(15)

But!

In forensics we are dealing with questioned signatures that could be the product of not only natural signing behaviour, but behaviour that is unnatural.

Again 

Disturbed

18.12.2015 15

(16)

Un-naturally written signatures

Closed Loop Control mechanism (CLC-mechanism)

– Un-natural signing behaviours are usually influenced by feedback

18.12.2015 16

(17)

Un-naturally written signatures

Closed Loop Control mechanism (CLC-mechanism)

– Un-natural signing behaviours are usually influenced by feedback

18.12.2015 17

(18)

Un-naturally written signatures

– Examples

18.12.2015 18

(19)

Un-naturally written signatures

– Further examples

18.12.2015 19

(20)

Un-naturally written signatures

– Further examples

18.12.2015 20

(21)

Un-naturally written signatures

– Further examples

18.12.2015 21

(22)

Un-naturally written signatures

– Further examples

18.12.2015 22

(23)

Summarizing the defects of simulations/forgeries

– Hesitation

– Un-natural pen lifts – Patching

– Uncertainity of movement – Stilted (drawn) quality

18.12.2015 23

(24)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 24

(25)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 25

(26)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 26

(27)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 27

(28)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 28

(29)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 29

(30)

What of Disguise

– Strategy

Exemplar Denial

18.12.2015 30

(31)

Predictors of disguise behaviour

– Presence of altered capital letters – Slower writing speed

– Faster writing speed

– Altered letter construction – Smaller letter heights

– Larger letter heights – Omitted letters in name – Added letters in name – Crowding of letters

– Stretched form (expanded) – Scrawling of form

– Slant alteration

– Use of crude letter forms – Altered terminal stroke

18.12.2015 31

(32)

Performance Analysis of FHEs

– On La Trobe signature data collection 2002 (9R, 20D, 104F, 76G) – Proficiency Tests

– provided with a scanned and printed hardcopy of each signature – provided additional info.

– recorded opinion on authenticity on a five-point scale – Authorship Group Results

– 8600 authorship opinions recorded – 5306 (61.7 %) were correct

– 60 (0.7 %) were incorrect/misleading – 3234 (37.6 %) were inconclusive – Overall error/misleading rate 1. 1%

– Important

– a lot of inconclusive cases

18.12.2015 32

(33)

Experts' difficulty: Inconclusive cases

Is that a problem for you as well? We’ll see;

3084

305 1917

0 184 37 23

518 2532

0 500 1000 1500 2000 2500 3000 3500

Genuine Disguised Simulated Questioned signature type

N um be r of op ini on s

# Correct

# Misleading

# Inc.

18.12.2015 33

(34)

Performance Analysis of FHEs

– On La Trobe signature data collection 2006 (25R, 7D, 90F, 3G) – Proficiency Tests

– provided with a hardcopy of each signature and an answer booklet – provided additional info.

– recorded opinion on authenticity on a five-point scale – Authorship Group Results

– 3100 authorship opinions recorded – 1254 (40.5 %) were correct

– 224 (7.2 %) were incorrect/misleading – 1622 (52.3 %) were inconclusive

– Overall error/misleading rate 15.2 % – Important

– a lot of inconclusive cases

– a lot of inter-expert variations exist

18.12.2015 34

(35)

Inter-expert variations

– Total number of misleading authorship opinions

– As can be observed, the total number of misleading opinions expressed on the trial varies. 10 response booklets contained no misleading opinions. The balance of the booklets contained between 1 and 34 misleading opinions.

18.12.2015 35

(36)

Experts' difficulty: disguise identification

Is that a problem for you as well? Lets see;

1151

0 96 113

93 10

111 0

1526

0 500 1000 1500 2000

Genuine Disguise Simulated

Questioned signature type

N u m b e r o f o p in io n s

# Correct

# Misleading

# Inc.

18.12.2015 36

(37)

Hands on Session: Real Case Work

Is that a problem for you as well? Lets see;

18.12.2015 37