The use of fingerprints available on the web in false identity documents: Analysis from a forensic intelligence perspective

Accepted Manuscript Title: The use of fingerprints available on the web in false identity documents: Analysis from a forensic intelligence perspective Author: Carlos Magno Alves Girelli PII: DOI: Reference:

S0379-0738(16)30063-9 http://dx.doi.org/doi:10.1016/j.forsciint.2016.02.041 FSI 8348

To appear in:

FSI

Received date: Revised date: Accepted date:

20-7-2015 28-1-2016 22-2-2016

Please cite this article as: C.M.A. Girelli, The use of fingerprints available on the web in false identity documents: analysis from a forensic intelligence perspective, Forensic Science International (2016), http://dx.doi.org/10.1016/j.forsciint.2016.02.041 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The use of fingerprints available on the web in false identity documents: analysis from a forensic intelligence perspective Carlos Magno Alves Girelli a,b,* a

Laboratory of Carbon and Ceramic Materials, Department of Physics, Federal University of Espirito Santo, 29075-910 Vitoria-

Identification Group, Federal Police Department of Brazil, 29114-670 Vila Velha-ES, Brazil

cr

b

ip t

ES, Brazil

us

* Correspondence to: Grupo de Identificação, Superintendência Regional de Polícia Federal no Espírito Santo, Rua Vale do Rio Doce, 01, São Torquato, Vila Velha-ES, CEP 29114-670, Brazil. Tel.: +55 27 3041 8089; fax: +55 27 3041 8064.

Ac ce p

te

d

M

an

E-mail address: [email protected] or [email protected]

Page 1 of 42

Abstract Fingerprints present in false identity documents were found on the web. In some cases, laterally reversed (mirrored) images of a same fingerprint were observed in

ip t

different documents. In the present work, 100 fingerprints images downloaded from the web, as well as their reversals obtained by image editing, were compared between

cr

themselves and against the database of the Brazilian Federal Police AFIS, in order to

us

better understand trends about this kind of forgery in Brazil. Some image editing effects were observed in the analyzed fingerprints: addition of artifacts (such as watermarks),

an

image rotation, image stylization, lateral reversal and tonal reversal. Discussion about lateral reversals’ detection is presented in this article, as well as suggestion to reduce

M

errors due to missed HIT decisions between reversed fingerprints. The present work aims to highlight the importance of the fingerprints’ analysis when performing

d

document examination, especially when only copies of documents are available,

te

something very common in Brazil. Besides the intrinsic features of the fingermarks considered in three levels of details by ACE-V methodology, some visual features of

Ac ce p

the fingerprints images can be helpful to identify sources of forgeries and modus operandi, such as: limits and image contours, fails in the friction ridges caused by excess or lack of inking and presence of watermarks and artifacts arising from the background. Based on the agreement of such features in fingerprints present in different identity documents and also on the analysis of the time and location where the documents were seized, it is possible to highlight potential links between apparently unconnected crimes. Therefore, fingerprints have potential to reduce linkage blindness and the present work suggests the analysis of fingerprints when profiling false identity documents, as well as the inclusion of fingerprints features in the profile of the documents.

Page 2 of 42

Keywords: Fingermark; False document; Web; Reversal; AFIS; Forensic intelligence.

ip t

1. Introduction

cr

According to the criminal database of the Brazilian Federal Police (BFP),

us

more than 7,000 cases involving false documents are investigated annually by that law enforcement agency [1]. Such statistics would become so much higher if considered the

an

criminal cases investigated by the 27 state polices of Brazil.

When persons bearing supposedly false documents are arrested, the seized

M

documents are sent to the forensic division to be examined. The analysis aim to verify

d

the authenticity of the document based mainly on features of the substrate. When the

te

criminal activity is detected after the offender has gone, generally only copies of the false documents are available for the investigation, resulting in the impossibility of

Ac ce p

conducting regular forensic analysis on the original questioned documents. Fingerprint examination may be helpful for detecting forgeries, especially when dealing with copies of documents, given that in Brazil identity (ID) documents usually display the fingerprint (right thumb) of the bearer. In late 2014, while searching on AFIS for a suspicious fingerprint present in

false ID document, experts from BFP in the state of Espirito Santo obtained HIT decisions with fingerprints from false documents that have been used in different states of Brazil, some of them located so far from the others. In general, no additional intelligence work was done with respect to the links created inside the AFIS by making

Page 3 of 42

HIT decisions. The results of the comparisons were solely focused on specific investigation process and the subsequent trial. Intrigued by the unlikely connection between those criminal cases from so

ip t

far apart locations, the experts decided to go beyond the routine task. They performed

search for the suspicious fingerprint image on the web, considered the most likely

cr

source for independent forgers have obtained a same fingerprint image. The search was

us

successful; they found websites containing that fingerprint used in multiple false documents [2]. The finding caused great surprise, given that all the experts of that

to search for fingerprints on the web before.

an

Identification Group had more than ten years of experience and they had never thought

M

The present study was designed to verify whether fingerprints available on the web besides that already identified have been used to forge ID documents in Brazil

d

and, if so, how often this has happened. Furthermore, considering that BFP has recently

te

detected some cases of laterally reversed (mirrored) fingerprints on false ID documents

Ac ce p

[2,3], and that such reversals have also been used in recent research of other authors [4], all procedures performed in the present study were applied for both fingerprints obtained from the web and their lateral reversals. Some image editing effects observed in the fingerprints obtained from the

web will be discussed with regard to their detection by the AFIS search. In this sense, a suggestion to improve the workflow when searching for lateral reversals on AFIS will be presented. Also, this web-based study will be analyzed from a forensic intelligence perspective. In simple terms, the intelligence activity can be briefly defined as the result of a process aiming to transform raw data into a form suitable for making decisions [5].

Page 4 of 42

Forensic intelligence occurs when such activity is carried out by law enforcement agencies in an accurate, usable and timely manner, obtaining information from traces and forensic case data to support tactical, operational and strategic decisions, especially

ip t

in models such as intelligence-led policing [6-8]. Instead of focusing on each individual criminal case and in the use of its evidences solely for court purposes, forensic

cr

intelligence is based on a multi-case focus and a broader approach. The aim is to uncover potential links that may lead to the identification of common sources and series

us

of crimes, allowing authorities to better understand crimes and use resources in a

an

proactive manner [8].

Some examples of the use of forensic intelligence are given elsewhere [5,9-

M

12]. Rossy et al. [9] have described the integration of retrospective dataset extracted from a common database shared by Swiss police forces, concluding that forensic

d

outcomes have a great potential to detect crime series. Morelato et al. [5] have discussed

te

the use of forensic case data in intelligence-led policing, presenting as example the illicit drug profiling performed in Australia and in Europe, in particular in Switzerland.

Ac ce p

A novel forensic intelligence model based on systematic profiling of false ID document was proposed by Baechler, Ribaux and Margot [10], aiming to uncover links, patterns and trends based on visual features of false documents. The application of such method to different types of seized documents has led those authors to conclude that it has a great potential to diminish linkage blindness and to develop analysis capacity at the strategic, operational and tactical levels [10]. In a follow-up research, Baechler et al. [11] have presented further results from the application of the profiling method to seized ID documents, giving more details about the comparison process and metrics used in the method. A transversal model comparing illicit drugs and false ID documents monitoring from a forensic intelligence perspective was presented by Morelato et al.

Page 5 of 42

[12], aiming to generalize the use of the method to break barriers between apparently separate fields of study in forensic science and intelligence, among other considerations. In Brazil, profiling of illicit drugs has been done [13-16], but there is still no

ip t

profiling of false documents for forensic intelligence purposes. Fingermarks and

documents are generally examined by different experts from different divisions and no

cr

forensic intelligence work has been developed with respect to fingerprints present on

us

false documents. One goal of this work is to present some features of fingerprints that have been used in false ID documents with potential to highlight possible links between

an

criminal cases, justifying the addition of fingerprint analysis in the profiling activity as

M

well as the inclusion of fingerprints features in the profile of the false ID documents.

te

d

2. Material and methods

Ac ce p

2.1. Web search and image editing

Fingerprints images were obtained by searching on Google Images website.

After typing ‘impressão digital’ (‘fingerprint’ in Portuguese language), the first displayed 100 fingerprint images in JPEG format were saved. Image resolutions ranged from 72 to 762 ppi, the most common incidences being 96 ppi (n = 49) and 300 ppi (n = 20). The images were edited using Adobe Photoshop in order to remove watermarks, texts, logo marks and any artifacts considered extrinsic to the fingerprints,

Page 6 of 42

such as for instance those displayed in Fig. 1. After clean up the images and orientate the fingerprints with the fingertips facing up, the samples were identified by numbers

M

an

us

cr

ip t

from ‘01’ to ‘100’ and are presented in Fig. 2.

Ac ce p

te

d

Fig. 1. Examples of artifacts present in fingerprints available on the web. The contrast between artifacts and fingerprints were highlighted using Adobe Photoshop for better visualization.

Page 7 of 42

# 02

# 03

# 04

# 05

# 06

# 07

# 08

# 09

# 10

# 11

# 12

# 13

# 14

# 15

# 16

# 17

# 18

# 19

# 20

# 21

# 22

# 23

# 24

# 25

# 26

# 27

# 28

# 31

# 32

# 33

# 34

# 35

# 36

# 37

# 38

# 39

# 40

# 41

# 42

# 43

# 44

# 45

# 47

# 48

# 49

# 50

# 51

# 52

# 53

# 54

# 55

# 56

# 57

# 58

# 59

# 60

# 61

# 62

# 63

# 65

# 66

# 67

# 68

# 69

# 70

us

an

# 30

cr

# 29

te

d

M

# 46

ip t

# 01

Ac ce p

# 64

# 71

# 72

# 73

# 74

# 75

# 76

# 77

# 78

# 79

# 80

# 81

# 82

# 83

# 84

# 85

# 86

# 87

# 88

# 89

# 90

# 91

# 92

# 93

# 94

# 95

# 96

# 97

# 98

# 99

# 100

Fig. 2. Sampling of 100 fingerprints images (edited) taken from the web.

Page 8 of 42

ip t

2.2. AFIS search

All the 100 fingerprints displayed in Fig. 2 were inserted, one by one, in the

cr

BFP AFIS. Because of the different sizes and resolutions of the images, it was

necessary to adjust the scales following an operational procedure developed by

us

fingermark experts from BFP [17]. The procedure consists in marking a distance

an

perpendicular to 11 ridge lines close to the nuclei region and attributes a length of 5 mm to that distance. Although the statistical study performed by those researchers [17] has

M

concluded that there are limitations associated to the method and it should not replace the use of a scale when photographing fingermarks, its use can be helpful as a last

te

the image.

d

resource for adjusting the scale in the lack of other reference with known dimension on

Ac ce p

The adjusted-scale fingerprints were auto-encoded by the system with no human manual mark-up of minutiae. A list of 15 candidates was presented by the system at the end of the search and HIT decisions were based not only on the number of minutiae, but on the whole set of features in agreement, as established by international standards [18-21]. Anyway, only for information, the numbers of minutiae in agreement for each pair of matched fingerprints were in the range 15-100. The first comparisons performed in AFIS were made between the fingerprints themselves. Matched fingerprints were grouped together in groups numbered by G-01, G-02, G-03 and so on. The group G-01 was filled with the fingerprint # 01 and all the others resulting from a HIT decision with fingerprint # 01; G-02 was filled with fingerprint # 02 (or the next in line if fingerprint # 02 have already

Page 9 of 42

been grouped) and all the others resulting from a HIT decision and so forth until all the 100 fingerprints being inserted in a group. After that, the 100 fingerprints were searched for tenprints (TP) and

ip t

unsolved latent prints (UL) from criminal cases. The searches were performed over the

whole database of the BFP AFIS, including not only criminal TPs and ULs from all

cr

types of crimes. According to the National Institute of Identification (INI), central unit

us

of BFP responsible for identification standards and criminal statistics, at the time of the experiments the database of the BFP AFIS was approximately 17 million tenprints (~

an

170 million fingerprints) and 170,000 latent prints (R.D. Cario from BFP AFIS management, personal communication, January 11, 2016).

M

In order to search for lateral reversals, all the original images taken from the web were laterally reversed using a computational tool available on the BFP AFIS

d

during the image acquisition stage. Then, they were codified following the same

te

methodology applied for the original fingerprints and all searches described above were

Ac ce p

performed again, including comparisons between original and reversed fingerprints. The groups formed by the laterally reversed fingerprints are referred to as G-01-LR, G-02LR and so on.

3. Results and Discussion

3.1. AFIS Search

Page 10 of 42

The results of comparisons between the fingerprints obtained from the web are presented in Table 1, with indication of the 56 formed groups and their

Table 1. Groups formed by matched fingerprints obtained from the web. G-02

G-03

G-04

G-05

G-06

G-07

02

03

04, 05, 09, 10, 15, 29, 45, 57, 77, 89

06

07, 26

08

G-11

G-12

G-13

G-14

G-15

G-16

G-17

14

16

17, 71

18, 64, 66

19

20

G-21

G-22

G-23

G-24

G-25

G-26

25, 48

28, 37, 53, 90, 96

30

31

32

33

G-31

G-32

G-33

G-34

G-35

G-36

41, 58

42, 75

44, 97

46, 83

47

49

G-41

G-42

G-43

G-44

G-45

60, 69

63

67

70

74

G-51

G-52

G-53

G-54

91

93

94

95

G-09

G-10

11, 52, 62, 65, 72

12, 56

13, 27

G-18

G-19

G-20

21, 43

22, 39, 68, 85

23

24, 35

G-27

G-28

G-29

G-30

34, 61

36

38, 55, 86

40

G-37

G-38

G-39

G-40

50, 73, 78, 87

51

54

59

G-46

G-47

G-48

G-49

G-50

76

79

80

82, 88, 92, 98

84

M

an

us

01, 81

G-08

cr

G-01

ip t

corresponding mated fingerprints.

G-56

99

100

d

G-55

te

Based on TP results the donor of the fingerprint presented twice in Group

Ac ce p

G-33 was identified. His tenprint had been previously acquired and inserted in AFIS at the moment he was getting a passport in a BFP station. For ethical reason, the donor’s identity is not revealed. No HIT decision from TP search was obtained for laterally reversed fingerprints.

The UL search resulted in 34 HIT decisions with fingerprints examined in

criminal cases, all of them related to false ID documents, being 31 positive results with fingerprints from the original groups and 3 with laterally reversed fingerprints. The groups containing such matched fingerprints are indicated in Fig. 3, as well as the number of correlated criminal cases for each group. It is worthy of mention that some criminal cases had many false ID documents such as, for instance, a case with 17 false

Page 11 of 42

documents displaying the same fingerprint matching fingerprints from group G-19. Therefore, the number of false ID documents presenting fingerprints supposedly taken

te

d

M

an

us

cr

ip t

from the web in this work is much larger than 34.

Ac ce p

Fig. 3. Number of correlated criminal cases to the groups of fingerprints taken from the web.

For the sampling of 100 original fingerprints taken from the web, 31

fingerprints belonging to 11 groups (G-02, G-04, G-08, G-19, G-20, G-22, G-31, G-39, G-41, G-42 and G-45) have matched fingerprints present on false documents. In terms of total number of fingerprints, it means that 31% of the fingerprints available on the web were detected as being used to forge ID documents in Brazil. This consideration is valid if we assume that the fingerprints printed on the false documents came from the web and not the opposite, what is a reasonable statement as will be discussed in section 3.3. This statistics is probably so much higher, given that only part of the committed crimes were investigated and only part of the investigated crimes were inserted in the

Page 12 of 42

BFP AFIS. Most of the crimes investigated by state polices are not included in the BFP AFIS. If we are interested only in different fingerprints, we have to consider the number of groups instead of the number of single fingerprint images. Indeed, 11 out of 56

ip t

different fingerprints available on the web appear in false ID documents, equivalent to 20%, which remains a considerable statistics. By way of illustration, Figs. 4 and 5

Ac ce p

te

d

M

an

us

60 and the website [22] where fingerprint # 60 was obtained.

cr

display, respectively, a false ID document containing fingerprint matching fingerprint #

Fig. 4. False ID document containing fingerprint matching fingerprint # 60.

Page 13 of 42

ip t cr us

an

Fig. 5. Website [22] where fingerprint # 60 was obtained.

M

The UL search for laterally reversed fingerprints resulted in HIT decisions for 6 out of 100 images, those belonging to the groups G-08-LR and G-19-LR. It means

d

that 6% of the fingerprint images or, again in terms of groups (2 out of 56), 4% of

te

different fingerprints available on the web were detected as having been used in its

Ac ce p

mirrored version to forge ID documents in Brazil. It is important to note that the fingerprints from groups G-08-LR and G-19-LR (obtained by reversing those from groups G-08 and G-19 respectively) were not found on the web in its non-reversed mode, at least for the considered sampling of 100 fingerprints. Therefore, such reversals used in false documents may have been either produced by the forgers or directly obtained from websites not included in the analyzed sampling. By way of illustration, Figs. 6 and 7 display, respectively, a copy of false ID document containing fingerprint matching fingerprint from group G-08-LR and a website [23] exhibiting fingerprint (not reversed) from group G-08.

Page 14 of 42

ip t cr us an

Ac ce p

te

d

M

Fig. 6. Copy of false ID document (with scale) containing fingerprint matching fingerprint from group G-08-LR.

Fig. 7. Website [23] exhibiting fingerprint (not reversed) from group G-08.

Page 15 of 42

The comparisons of fingerprints from the original groups against the reversed ones also resulted in HIT decisions, indicating that some fingerprints can be found on the web in both original and laterally reversed versions. Four groups were

ip t

identified as being lateral reversals of other four groups, but it is not possible to know which ones have original or reversed prints, since their sources are unknown. The four

cr

pairs of groups containing laterally reversed fingerprints are G-09 ↔ G-13, G-20 ↔ G22, G-39 ↔ G-52, and G-46 ↔ G-49. These additional links reduce the total number of

us

groups from 56 to 52 groups of fingerprints originated from the same source (donor), and the statistics cited above become even higher. By way of illustration, Figs. 8 and 9

an

display, respectively, website [24] exhibiting fingerprint # 56 from group G-09 and

Ac ce p

te

d

are lateral reversal of each other.

M

website [25] exhibiting fingerprint # 17 from group G-13. Fingerprints # 56 and # 17

Fig. 8. Website [24] exhibiting fingerprint # 56 from group G-09, lateral reversal of fingerprint # 17 from group G-13 (displayed in Fig. 9).

Page 16 of 42

ip t cr us an

M

Fig. 9. Website [25] exhibiting fingerprint # 17 from group G-13, lateral reversal of fingerprint # 56 from group G-09 (displayed in Fig. 8).

te

d

3.2. Image editing observed in the fingerprints and their effects on AFIS search

Ac ce p

Most of the assessed fingerprints have presented some kind of image editing. The main types of image editing that have been observed in the images are, by order of higher incidence: addition of artifacts, image rotation, image stylization, lateral reversal and tonal reversal. Description of such image editing and the fingerprints in which they were observed are presented in Table 2.

Table 2 Description of the main types of detected image editing and identification of the fingerprints in which they were observed. Image editing

Description

Fingerprint #

Addition of artifacts

Presence of watermarks, symbols or texts on the image and its surrounding.

Not observed for: 6, 14, 16, 18, 19, 30, 32, 33, 41, 42, 46, 48, 50, 51, 53, 54, 55, 58, 71, 73, 74, 75, 76, 78, 79, 81, 82, 83, 84, 90, 98, 100.

Page 17 of 42

Image position different from the conventional where fingertips are facing up.

6, 12, 13, 15, 21, 24, 27, 33, 35, 43, 56, 57, 63, 77, 87, 90, 94.

Image stylization

Image transformation over the friction ridge shape, after which the image becomes, generally, more friendly, such as in cartoons.

15, 17, 25, 34, 38, 48, 49, 57, 61, 67, 86, 98.

Lateral reversal*

Rotation of 180̊ around an axis contained in the plane of the fingerprint, giving rise to a mirrored image of the original print.

17, 24, 35, 71, 76, 93.

Tonal reversal

Color inversion (black ↔ white).

87, 88, 89.

ip t

Image rotation

us

cr

* Fingerprints in larger quantities were assumed as references (not reversed). For groups containing the same amount of reversed prints, the reference was fixed on the fingerprint displayed first in the list of samples (Fig. 2).

Except for lateral reversals, the types of image editing listed in Table 2 are

an

relatively easy to detect and to correct. A brief discussion about these observed image

M

editing and the effects on AFIS search is given below.

Artifacts added to fingerprints images, when presenting good contrast can

d

be easily noticed. In such cases, the examiner may decide whether to clean up the image

te

using computational resources or perform AFIS encoding and remove eventual false minutiae due to the artifacts. Artifacts presenting low contrast will probably not

Ac ce p

considered by AFIS and no false minutiae will be auto-encoded by the system. Rotation problems can also be easily detected and solved. In most cases,

fingermarks can be oriented properly based on the ridge flow and anchor points such as deltas and cores. In the lack of such elements, for instance for small marks or low quality images, AFIS systems usually offer the possibility to perform a complete 360° search for disoriented fingermarks. In this work, issues regarding orientation and artifacts were soon detected and solved before AFIS input, as described in Section 2.1. Detection of image stylization depends on how deep was the intervention on the image. Different types of image stylization were observed. Some examples are

Page 18 of 42

displayed in Fig. 10. Perhaps some of the images have not been originated from image editing of a given fingerprint, but may be created by computer.

However, no

assumption was made in this sense and all images were equally treated in this work. A

ip t

good example showing that we should not discard such stylized images is given in Fig. 11. The stylized fingerprint # 15 from group G-04 is displayed, exactly as saved from

cr

the web, side-by-side with a matched fingerprint used in false ID document investigated

Ac ce p

te

d

M

an

us

by BFP.

Fig. 10. Some image stylization observed in the fingerprints images available on the web.

Page 19 of 42

ip t cr us an

M

Fig. 11. Fingerprint # 15, as saved from the web, presenting addition of artifacts, image stylization and image rotation, and a matched fingerprint present in copy of false ID document investigated by BFP.

d

Tonal reversal occurs when the color of ridges and furrows are inverted [26-

te

29]. In difficult cases where distinction between ridges and furrows is not evident, the

Ac ce p

examiner can observe some clues to know that the print is tonally reversed [29]. In this present work tonal reversals were easily detected because of their dark background. Lateral reversal is originated when the fingermark is rotated 180° around an

axis located in the plane of the fingermark, giving rise to a mirrored image of the original mark. Some cases of laterally reversed fingermarks have been previously reported in the literature [2,3,30-33]. The reversed marks can arise: (1) unintentionally, from the contact between two surfaces, transferring the fingermark deposits present on a touched surface to another, or (2) intentionally, by means of the use of image editing software, as can be deduced for the laterally reversed fingerprints observed in this work.

Page 20 of 42

Different from the other types of image editing discussed previously, lateral reversals cannot be detected only by looking at the fingerprints. Unless the examiners have strong suspicious about someone or the circumstances where fingermarks were

ip t

developed indicate possible ridge skin matrix transfer, there are no clues for help examiners in detecting the occurrence of such reversal. In general, the regular search on

cr

AFIS for a given fingermark will not display the reversal in the list of most likely candidates, because of the difference in the minutiae map. While for image stylization

us

and tonal reversal the contrast between ridges and furrows remains at the same position and the minutiae map is the same of the original fingermark, for lateral reversals the

an

map becomes inverted. Fig. 12 displays the minutiae maps obtained by auto-encoding

M

in AFIS for fingerprints # 72, # 82, # 88 and # 98. Defining fingerprint # 82 as reference, fingerprint # 72 is its lateral reversal, fingerprint # 88 its tonal reversal and

d

fingerprint # 98 its stylized image. The overlap of images indicates agreement for

te

minutiae maps of fingerprints # 82, # 88 and # 98, and disagreement for fingerprints #

Ac ce p

82 and # 72, as stated before.

Page 21 of 42

ip t cr us an M d te Ac ce p

Figure 12. Overlapping of minutiae maps obtained by auto-encoding in AFIS, demonstrating agreement for fingerprints # 82 (assumed as reference), # 88 (tonal reversal) and # 98 (stylized image), and disagreement for fingerprints # 82 and # 72 (lateral reversal).

Therefore, to be found, lateral reversals must be searched, and here arises the issue about the involved costs and benefits. Previous work [3] has suggested the search for lateral reversals when dealing with supposedly fake documents. Experts from BFP in the state of Espirito Santo have searched for reversals in criminal cases

Page 22 of 42

involving supposedly fake ID documents, whenever no suspect is identified by TP search. In 2015, 111 criminal cases were investigated by that Identification Group. The suggested standard procedure was followed in 8 of those criminal cases, containing a

ip t

total of 15 fingerprints present in questioned ID documents. The UL search for lateral reversals resulted in 3 HIT decisions (3 different analyzed cases), establishing links with

cr

other criminal cases that would not be linked if only conventional search for the original fingerprints was done. Therefore, the application of the standard procedure resulted in

us

HIT decisions for 38% of the analyzed cases (3 out of 8), or 20% (3 out of 15) of the total amount of fingerprints where lateral reversals were searched. Although such

an

statistics is based on a limited number of cases, it suggests that the benefits have

M

justified the costs in that particular reality. Stakeholders must evaluate the cost-benefit relationship for the application of such standard procedure based on their own

d

parameters and reality.

te

Also regarding the search for lateral reversals in AFIS, we have a suggestion for improving the system in order to optimize the workflow. There are many different

Ac ce p

AFIS algorithms in market nowadays. The one used in this work has a tool to laterally reverse images during image acquisition, the first stage of the process. Probably such a tool was thought and made available to correct reversed image position of fingermarks arising from transparent substrates, lifts, deposited with tape or directly photographed from the finger. When performing search for reversals, examiners must follow the same procedure used for original fingermarks. The image is acquired, edited, codified and so forth. The expert must repeat the whole procedure twice, one for the original questioned fingermark and other for its reversal. AFIS systems can be improved to enable search for fingermarks and their reversals without requiring double working. This can be done, for example, considering

Page 23 of 42

that the minutiae map of the reversal can be deduced from the minutiae map of the questioned fingermark. Figure 13 displays a fingermark in the first quadrant (right) and

M

an

us

cr

ip t

its reversal in the second quadrant (left) of the same framework.

te

d

Figure 13. Possible relationship between minutiae of reversed fingermarks.

Ac ce p

For each minutia with coordinates (x, y, θ) in the original fingermark corresponds an equivalent minutia with coordinates (-x, y, - θ) in its reversal. The minutiae map of the reversed fingermark can be derived from the minutiae map of the original one without the need of a new codification. Therefore, instead of performing twice the whole process perhaps it may be possible to improve the AFIS algorithm to offer the possibility to search for reversal after the codification process, given that image editing and codification are the same for both the original and the reversed fingermarks. This could make great difference, especially when fingermarks present low quality, such as for latent marks arising from crime scenes or even for fingerprints present on bad copies of ID documents as some used in this work. In such cases where the number of features used for comparison approach the threshold for an identification

Page 24 of 42

decision, besides save time and consequently money, the implementation of the suggested improvement in the system could result in more reliable results, given that the examiner would perform the hard task only once, reducing the risk of making mistakes

us

an

3.3. Analysis from a forensic intelligence perspective

cr

Professional system analysts will certainly present smarter solutions.

ip t

due to fatigue. The above mentioned correlation between minutiae is just an example.

The results obtained from AFIS search presented in section 3.1 can be used

M

not only to support the traditional investigative process focused on the trial, but also to support forensic intelligence. Potential links between cases can be highlighted by

d

observing the use of the same fingerprints (or some fingerprints’ features) in documents

te

seized at different places and times. Fig. 14 displays a Brazilian map, with the locations

Ac ce p

where the false documents containing fingerprints supposedly taken from the web were seized. The criminal cases are identified in the map according to the groups of fingerprints they have matched.

Page 25 of 42

ip t cr us an M d te Ac ce p

Figure 14. Brazilian map indicating the locations where the documents were seized. Symbols identify the corresponding groups of fingerprints available on the web.

According to Fig. 14, false documents containing fingerprints supposedly

taken from the web have been seized along the entire extension of the country. However, it is possible to observe concentration of seized documents in some regions. For instance, documents containing fingerprints from group G-08 are concentrated in central region and documents containing fingerprints from group G-41 are located in southeast region. The geographic distribution of such seizures suggests that those false

Page 26 of 42

documents may have been produced in their respective regions. Such presumption is in agreement with further information provided by the investigation, including the testimonies of some people who were arrested using those false documents. Indeed,

ip t

matched fingerprints in false ID documents can be considered as a valid option to help in detecting possible links between criminal cases.

cr

The forensic intelligence activity shall consider the whole available dataset

us

in order to reduce the risk of making erroneous decision. Returning to the example cited above involving false documents with fingerprints from group G-08, it is possible to

an

observe from Fig. 14 that one of the cases is located out of the central region where all the other cases were investigated. In principle, there is no obstacle for someone to get a

M

false document in the central region and to use it in the southeast region. However, it is also possible that such document seized in the southeast region had been produced by a

d

different forger in southeast region, since that fingerprint was available on the web to

te

anyone. Fig. 15 displays a timeline, indicating approximately the time when the documents were seized. The dashed red line indicates the web search’s date (December

Ac ce p

17, 2014). It is possible to see that one criminal case correlated to group G-08 is temporally distant from the others, having occurred almost two years before. Although the cases are not individually identified in Figs. 14 and 15 (they are only identified by the correlated groups), the case distant from the others in both graphs is, in fact, the same. Thus, in this example the time analysis has corroborated information resulting from the spatial analysis, strengthening the hypothesis that such considered case is not linked to the others.

Page 27 of 42

ip t cr us an

M

Fig. 15. Time distribution of crimes involving false ID documents containing fingerprints that have matched fingerprints taken from the web. The symbols identify the groups of the correlated fingerprints. The dashed red line indicates the web search’s date (December 17, 2014).

d

Another interesting example refers to false ID documents containing

te

fingerprints of the group G-39. According to Fig. 14, the 4 cases are spatially grouped in the central region, suggesting a common source in that region. On the other hand,

Ac ce p

Fig. 15 indicates that those 4 cases are temporally well grouped into two pairs, being one pair related to seizures made in 2008 and the other pair of cases to documents seized in 2014. Perhaps this time gap between seizures of documents containing the same fingerprints can be explained by considering the existence of two sources of forgery instead of one, since again those fingerprints could be available on the web to anyone. Conclusion in this way requires information beyond those available to this research. The analysis of Fig. 15 indicates a possible trend on the use of fingerprints taken from the web to forge false documents. According to Fig. 15, the criminal cases identified in this work have happened near the date on which the fingerprints were

Page 28 of 42

downloaded from the internet. At that time, those images were at the ‘top list’ presented by the Google search. Except for a pair of cases (G-39) containing documents seized in 2008, no other case older than 3 years (from the web search’s date) has presented some

ip t

of the studied fingerprints taken from the web. According to the BFP criminal database [1], the number of crimes involving false documents in Brazil was approximately

cr

constant along the last 10 years. Thus, there is no reason a priori to justify the

concentration of cases in the last years as displayed in Fig. 15. The key answer for this

us

issue seems to be the availability of the fingerprints on the web and the dynamic way in which such availabilities change over time. In this sense, this possible trend suggests

an

that fingerprints downloaded from the web and used in false documents could indicate

M

approximately when the documents were produced. As result, a systematic profiling of false ID documents could be benefited from a parallel acquisition of fingerprints images

d

from the web, helping to understand the behavior of the images’ availabilities over time.

te

The possible trend under discussion is based on the premise that false ID documents are used as soon they are forged, something to be proved, perhaps performing this same

Ac ce p

research in a few years.

The image analysis of fingermarks may result in additional conclusions

beyond the identification of the fingermarks’ sources [34]. Besides the intrinsic features of the fingermarks considered in three levels of details by the ACE-V methodology [18], there are also extrinsic features in fingermarks related to the substrate and conditions in which the marks were deposited. Some examples of extrinsic features are limits and image contours, fails in the friction ridges caused by excess or lack of inking and presence of watermarks and artifacts arising from the background. Because of their random and unpredictable nature, such features are not entirely reproduced in fingermarks obtained by distinct deposition. Thus, a qualitative and quantitative

Page 29 of 42

analysis of such features may help examiners to reach additional conclusion from fingermarks examination. Although there is no scientific basis to ensure unequivocal conclusion from such examination, given that extrinsic features do not follow the same

ip t

principles of the friction ridges (i.e. permanence and uniqueness [35]), this resource deserves to be considered at least as a subsidiary option, even if only for interna

cr

corporis procedures. Following are presented some application to fingerprints assessed

us

in this work.

Returning again to the cases with false ID documents presenting fingerprints

an

from group G-08, the analysis of the intrinsic features has led to conclude that the prints came from the same source (donor). In Fig. 16 are compared 4 fingerprints: fingerprint

M

# 52 from group G-08, and fingerprints present in 3 false ID documents arbitrarily labeled “A”, “B” and “C”. Documents “A” and “B” were seized in the central region in

Ac ce p

te

d

2014 and 2015, and document “C” was seized in the southeast region in 2012.

Fig. 16. Comparison between fingerprint # 52 from group G-08 and fingerprints present in 3 false ID documents (arbitrarily labeled “A”, “B” and “C”) from different criminal cases, all originated from the same source (donor). Based on the extrinsic features, it is possible to conclude that, except for False ID doc. “C”, the remaining fingerprints are copies of a same image. Red arrows indicate dots on the limits of the fingerprints, green circles indicate some points presenting excess of inking and blue circles/ellipses indicate lack of inking.

Page 30 of 42

Some concepts such as ‘details in agreement to identify’ and ‘details in disagreement to exclude’ [18], well-known for fingermark examiners, can be easily imported from the ridgeology to be applied in this process. In Fig. 16, some features in

ip t

agreement between the first three images are highlighted, suggesting that those fingerprints are copies of the same image. In contrast, the fingerprint present in

cr

document “C” is not a copy of the other 3 images. Such conclusion corroborates the

results from the time and spatial analysis, strengthening even more the possible lack of

us

linkage between the criminal case of document “C” and the others.

an

Also regarding this last example, it is possible to reach another interesting conclusion from the analysis of extrinsic features of the fingerprints. While the

M

fingerprints on false ID documents “A” and “B” were printed over the frame and their bottom part became overlapped with the frame, the fingerprint # 52 available on the

d

web is clean, free of such overlapping. As a consequence, it is more likely those

te

fingerprints had been taken from the web than the opposite, as previously stated in section 3.1. In addition, the action of printing the fingerprints on the frame rather than

Ac ce p

centralized within the appropriate space can be an important feature, indicative of the modus operandi, thus worthy of consideration in the profile of the document. In some cases was possible to detect watermarks on the fingerprints used in

false ID documents leading experts to promptly identify the falsification. Fig. 17 displays a false ID document containing fingerprint (group G-19) where an evident watermark (highlighted in black) was found, as well as the website [35] from where the fingerprint image was probably downloaded.

Page 31 of 42

ip t

us

cr

Fig. 17. False ID document containing fingerprint (group G-19) with visible watermark (highlighted), and the website [35] from where the fingerprint image was probably downloaded.

an

The same watermark displayed in Fig. 17 is present (in different sizes) in 25 out of the 100 analyzed fingerprints. These 25 fingerprints belong to 18 groups of

M

fingerprints, 6 of them containing fingerprints identified in false ID documents (G-04, G-08, G-19, G-20, G-41 and G-42). Other watermarks and artifacts were found in

d

several fingerprints considered in this research, some of them also used in false ID

te

documents. Even in low quality copies of documents was possible to see watermarks. Thus, the presence of such artifacts and the identification of the corresponding websites

Ac ce p

from where the fingerprints may have been downloaded seem to be relevant information worthy of consideration.

The spatial and temporal representations of criminal cases displayed in Figs.

14 and 15 respectively were based on the premise that documents sharing matched fingerprints can be potentially linked. It is also plausible to suppose that a given forger could use different fingerprints taken from the same website. In this sense, documents presenting different fingerprints, but fingerprints with the same watermark, available at the same website, could be also potentially linked. Spatial and temporal representations based on such linkage could also provide valuable information for forensic intelligence purposes, but this suggestion is not realized in this work.

Page 32 of 42

Therefore, according to the discussion presented above, some features of the fingerprints used in false ID documents have potential to highlight possible links between criminal cases. The present work suggests the analysis of fingerprints when

ip t

profiling false ID documents, including relevant features of the fingerprints in the profile of the documents. To be considered relevant, such features must be

cr

representative of their sources and of the modus operandi. The ability of these features

for representing the inter- and intra-sources variations can be evaluated through

us

different metrics and making use of known samples [10,11].

an

This research was developed using a limited number of samples and limited information about the criminal cases involving false ID documents. The confirmation of

M

possible links and trends that have been raised in this research, as well as the identification of others not cited, requires further work.

d

In consequence of the results obtained in this study, some actions are being

te

taken by the BFP. There is a work in progress in which hundreds (if not thousands) of

Ac ce p

fingerprints images taken from the web are systematically being inserted and stored in AFIS under specific numbering. Fingermark experts across the country will be able to make HIT decisions with fingerprints available on the web without the need to perform search on the web, which would mean an exhausting task, if it was done. After making a HIT decision, the reference number displayed by AFIS can be used to get additional information in the BFP intranet. The examiner will have access to a ‘print screen’ of the website from where the identified fingerprint was downloaded and further information such as link to the webpage and date of access, allowing the reporting of the results in a standardized way.

Page 33 of 42

Besides that, a pilot project to profile false ID documents is being considered by BFP in the state of Espirito Santo. It will be a good opportunity for document and fingermark experts work together in order to extract as much as possible

ip t

information from questioned documents. Depending on the results, the project can be expanded and applied at national scale, giving rise to a more complete and accurate

cr

diagnosis of the problem, aiming to provide more reliable information to support the

us

authorities in decision-making.

M

an

Conclusion

Fingerprints present in questioned ID documents can help in detecting

d

forgeries, especially when only copies of the documents are available for examination.

te

The analysis of fingerprints available on the web has demonstrated that they are

Ac ce p

frequently used to forge documents in Brazil. Many fingerprints obtained from the web have presented some type of

image editing, the more problematic being the lateral reversals, given the difficulty associated to its detection. The application of a standard procedure by the BFP in the state of Espirito Santo has presented a satisfactory cost-benefit ratio, indicating the recurrent use of laterally reversed fingerprints in false ID documents. Aiming to optimize the search for reversals, the present study suggests a technological implementation in AFIS systems, taking advantage of the relationship between reversed images. By this improvement, the expert could search for both original and reversed images by encoding the fingerprint only once, resulting in economy and celerity, as well as reducing the risk of error caused by fatigue.

Page 34 of 42

From a forensic intelligence perspective, fingerprints present in false ID documents have potential to highlight possible links between different criminal cases and trends regarding the use and making of false ID documents. The analysis and

ip t

comparison of intrinsic and extrinsic features of the fingerprints allow examiners to reach important conclusions, such as for instance, that the fingerprints: came from the

cr

same source (finger); are copies of a same image; were taken from the web and not the

opposite; were downloaded from the same website; are (were) available on the web in a

us

given time; were image edited. The spatial and time analysis of seizures of false ID documents sharing these features, together with the whole available dataset of the

an

investigation, can provide relevant information to support authorities in decision-

M

making at different levels.

Therefore, the present work suggests the analysis of fingerprints when

d

profiling false ID documents, as well as the inclusion of fingerprints features in the

te

profile of the documents. Some actions are being taken by BFP in this sense, including

Ac ce p

systematization of procedures and documents profiling.

References

[1] SINIC - Sistema Nacional de Informações Criminais. Ministério da Justiça – Departamento de Polícia Federal, BRASIL. Accessed January 6, 2016. [2] C.M.A. Girelli, Application of a standard procedure to avoid errors when comparing fingerprints with their reversals in fake documents. J. Forensic Sci. Med.

Page 35 of 42

[Epub

ahead

of

print]

Available

from: http://www.jfsmonline.com/preprintarticle.asp?id=162807 (accessed on 4.1.16). [3] C.M.A. Girelli, Laterally reversed fingerprints detected in fake documents, J.

ip t

Forensic Identif. 65 (1) (2015) 1-17.

cr

[4] G. Langenburg, C. Hall, Q. Rosemarie, Utilizing AFIS searching tools to reduce errors in fingerprints casework. Forensic Sci. Int. 257 (2015) 123-133.

us

[5] M. Morelato, A. Beavis, M. Tahtouh, O. Ribaux, P. Kirkbride, C. Roux, The use of

an

forensic case data in intelligence-led policing: The example of drug profiling, Forensic Sci. Int. 226 (2013) 1-9.

M

[6] O. Ribaux, A. Girod, S.J. Walsh, P. Margot, S. Mizrahi, V. Clivaz, Forensic intelligence and crime analysis, Law Probab. Risk 2 (2003) 47-60.

d

[7] O. Ribaux, S.J. Walsh, P. Margot, The contribution of forensic science to crime

te

analysis and investigation: Forensic intelligence, Forensic Sci. Int. 156 (2006) 171-181.

Ac ce p

[8] J.H. Ratcliffe (Ed.), Intelligence-led Policing, Willan Publishing, Cullompton, 2008. [9] Q. Rossy, S. Ioset, D. Dessimoz, O. Ribaux, Integrating forensic information in a crime intelligence database, Forensic Sci. Int. 230 (2013) 137-146. [10] S. Baechler, O. Ribaux, P. Margot, Toward a novel forensic intelligence model: systematic profiling of false identity documents, Forensic Sci. Policy Manag. Int. J. 3 (2) (2012) 70-84. [11] S. Baechler, V. Terrasse, J.-P. Pujol, T. Fritz, O. Ribaux, P. Margot, The systematic profiling of false identity documents: method validation and performance evaluation

Page 36 of 42

using seizures known to originate from common and different sources, Forensic Sci. Int. 232 (2013) 180-190. [12] M. Morelato, S. Baechler, O. Ribaux, A. Beavis, M. Tahtouh, P. Kirkbride, C.

ip t

Roux, P. Margot, Forensic intelligence framework - Part I: Induction of a transversal model by comparing illicit drugs and false identity documents monitoring, Forensic Sci.

cr

Int. 236 (2014) 181-190.

us

[13] E.K. Shibuya, J.E.S. Sarkis, O.N. Neto, M.Z. Moreira, R.L. Victoria, Sourcing Brazilian marijuana by applying IRMS analysis to seized samples, Forensic Sci. Int. 160

an

(2006) 35-43.

M

[14] A.R. Fukushima, V.M. Carvalho, D.G. Carvalho, E. Diaz, J.O.W.V. Bustillos, H.S. Spinosa, A.A.M. Chasin, Purity and adulterant analysis of crack seizures in Brazil,

d

Forensic Sci. Int. 243 (2014) 95-98.

te

[15] M.C.A. Marcelo, K.C. Mariotti, M.F. Ferrão, R.S. Ortiz, Profiling cocaine by

Ac ce p

ATR–FTIR, Forensic Sci. Int. 246 (2015) 65-71. [16] S.F. Lapachinske, G.G. Okai, A. Santos, A.V. Bairros, M. Yonamine, Analysis of cocaine and its adulterants in drugs for international trafficking seized by the Brazilian Federal Police, Forensic Sci. Int. 247 (2015) 48-53. [17] C.E.P. Carvalho, M.O. Amaral, R.C.C. Valdes, M.M.V. Alecrim, Contagem com estatística para o NID. Revista Impressões. 7 (1) (2007) 20-29. [18] Scientific Working Group on Friction Ridge Analysis, Study and Technology (SWGFAST), Document #10 Standards for Examining Friction Ridge Impressions and Resulting Conclusions (Latent/Tenprint). Ver. 2.0. September 13, 2013. Available from:

Page 37 of 42

http://swgfast.org/documents/examinations-conclusions/130427_ExaminationsConclusions _2.0.pdf (accessed on 22.4.14). [19] Fingerprint examination: developing a quality standard. Forensic Science

ip t

Regulator. Research and Analysys. First published: 21 December 2001. Available from:

ev-quality-std-fingerprint-exam.pdf (accessed on 18.1.16).

cr

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/118947/d

August

1,

us

[20] IAI Resolution 1973-5. Report of the Standardization Committee of the IAI – 1973.

Available

from:

an

https://www.theiai.org/member/resolutions/1973/Resolution_1973-5.pdf (accessed on

M

18.1.16).

[21] IAI Resolution 2010-18. Change official position related to Friction Ridge

Available

from:

te

1980-5).

d

Examinations based on science and scientific research (rescinds Resolutions 1979-7 &

https://www.theiai.org/member/resolutions/2010/Resolution_2010-18.pdf (accessed on

Ac ce p

18.1.16). [22]

Available

from:

https://www.google.com.br/search?q=impress%C3%A3o+digital&newwindow=1&site =webhp&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj98Nzhx6rKAhWEiJAKHQ QeDEUQ_AUIBygB&biw=1366&bih=667#imgrc=heJtjH_sA5uVGM%3A

(accessed

on 14.1.16). [23]

Available

from:

https://www.google.com.br/search?q=fingerprint&biw=1600&bih=795&site=webhp&s ource=lnms&tbm=isch&sa=X&sqi=2&ved=0ahUKEwixjvfNnafKAhWKFZAKHYzlA FYQ_AUIBigB#imgrc=iG6zoNFkG8_wZM%3A (accessed on 13.1.16). [24]

Available

from:

https://www.google.com.br/search?q=impress%C3%A3o+digital&newwindow=1&site

Page 38 of 42

=webhp&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj98Nzhx6rKAhWEiJAKHQ QeDEUQ_AUIBygB&biw=1366&bih=667#imgrc=tJlV65lQy0iyaM%3A (accessed on 14.1.16).

cr

ip t

[25] Available from: https://www.google.com.br/search?q=impress%C3%A3o+digital&newwindow=1&site =webhp&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj98Nzhx6rKAhWEiJAKHQ QeDEUQ_AUIBygB&biw=1366&bih=667#imgrc=j30R3nfUgfRpgM%3A (accessed on 14.1.16). [26] A.J. Lowe, Recognition and identification of reverse color latent prints, J. Forensic

us

Identif. 38 (5) (1988) 227–230.

an

[27] D.S. Pierce, Tonally reversed friction ridge prints on plastics, J. Forensic Identif. 39 (1) (1989) 11–22.

M

[28] T. Sethman, Latent print tonal reversal: Case and effect, The Print 20 (3) (2004) 3– 4.

te

d

[29] S. Castellon, Clues in friction ridge comparisons: Tonal reversals, J. Forensic Identif. 64 (3) (2014) 223-237.

Ac ce p

[30] P.A. Lane, M. Hilborn, S. Guidry, C.E. Richard, Serendipity and super glue: Development of laterally reversed, transferred latent prints, J. Forensic Identif. 38 (6) (1988) 292-294.

[31] J. Saviano, The significance of using level 1 detail in latent print examinations, J. Forensic Identif. 53 (2) (2003) 209-218. [32] M.H. Kershaw, Laterally reversed, J. Forensic Identif. 50 (2) (2000) 138-140. [33] E.R. Czarnecki, Laterally inverted fingerprints, J. Forensic Identif. 55 (6) (2005) 702-706.

Page 39 of 42

[34] C.M.A. Girelli, Fingermarks: beyond the source, J. Forensic Identif. Manuscript accepted. [35] C. Champod, P. Margot, C. Lennard, M. Stoilovic, Fingerprints and other ridge skin impressions, CRC Press, Boca Raton, 2004. Fingerprint

Stock

Photos,

Images,

&

Pictures,

Available

from:

ip t

[36]

www.dreamstime.com/photos-images/fingerprint.html#details53974044 (accessed on

Ac ce p

te

d

M

an

us

cr

20.1.16).

Page 40 of 42

Highlights

• Fingerprints taken from the web have been used to forge ID documents in Brazil.

ip t

• Lateral reversals are difficult to detect and have been observed in false ID documents. • Fingerprints can be representative of their sources of forgery and modus operandi.

• Fingerprints present in false documents have potential to highlight links between

cr

cases.

Ac ce p

te

d

M

an

us

• The analysis of fingerprints is recommended when profiling false ID documents.

Page 41 of 42

Acknowledgements

Ac ce p

te

d

M

an

us

cr

ip t

The author gratefully acknowledges the fingermark experts from the Brazilian Federal Police for the information provided to support this study and the anonymous reviewers for the valuable comments which have greatly enriched this work.

Page 42 of 42