Authentication of Medicine Blister Foils: Characterization of the

Rotogravure Printing Process

Iuliia Tkachenko, Alain Tr

emeau and Thierry Fournel

Laboratoire Hubert Curien, UMR CNRS 5516, Universit

e de Lyon, UJM-Saint-Etienne,

18 rue Professeur B. Lauras, 42000 Saint-Etienne, France

Keywords:

Rotogravure Printing, Medicine Blister Authentication.

Abstract:

Nowadays the number of medicine packaging counterfeits increases very quickly. The rotogravure printing

technique is worldwide used for medicine blister foils production. However, the existing anti-counterfeiting

solutions do not take into account this printing process. Additionally, it is not easy to apply conventional

solutions while using blister foils instead of uncoated/coated paper. In this paper, we study some features of

the rotogravure printing and identify the future paths to ﬁght the increasing number of counterfeited medicine

products. We present the result of a preliminary study of such a process, extended to foils, and discuss some

promising solutions for blister foils authentication.

1 INTRODUCTION

The worldwide market is suffering from packaging

counterfeiting. According to the Association for

Packaging and Processing Technology

, such a threat

is predicted to increase three percent per year world-

wide. Medical supplies represent one of the most sen-

sitive markets: counterfeiting affects health of com-

mon people causing damage of brand reputation and

market loss.

One of the ﬁrst solutions for the authentication

of printed surfaces was proposed by Goldman et al.

(Goldman, 1983), making use of paper ﬁbers and en-

graved dots as measurable but not duplicable physical

characteristics. The use of individual printable glyph

or character as a forensic mark were later investiga-

ted in a way consisting in the extraction of a proﬁle

of the printing: in (Pollard et al., 2010), a Model Ba-

sed Signature Proﬁle (MBSP) is extracted during the

printing process, then stored in the database. The sto-

red MBSP is compared with MBSP extracted from

document during veriﬁcation using the Shape Distor-

tion Encoding Distance (SDED), a modiﬁed Ham-

ming distance.

In (Kee and Farid, 2008), a proﬁle of the legal

printer is learnt on samples of a given character via

principal component analysis during the registration

step. At veriﬁcation, this proﬁle is used to assess

PMMIs full 2016 Brand Protection and Product Trace-

ability report https://www.pmmi.org

the link of the current document with the printer by

computing the error of the reconstruction of the new

occurrences of the character. Investigation whether

the document was printed using a particular device

(Navarro et al., 2018) is a second approach of prin-

tings authentication.

Instead of using a model for legitimate prints or

printings, another approach was suggested: the as-

sessment of the degradation caused by printing an

anti-copy pattern specially designed to protect the do-

cument or packaging (Picard, 2004). The so-called

Copy Detection Patterns (CDPs) are maximum en-

tropy images generated using a secret key and derived

from the content, that takes full advantage of the prin-

ciple of information loss during the printing process.

A quality index is here considered to measure the non

duplicable impact of the process.

A modular version was proposed for protecting

QR codes against copy including error correction for

keeping their reading robust at a second level (Tka-

chenko et al., 2016). The ﬁrst level of these ”2LQR”

codes is accessible to all public and can be read by

standard barcode reader. The second level is accessi-

ble only for authorized users and is sensitive to du-

plication process. The copy sensitivity is ensured

by the use of speciﬁc textured patterns sensitive to

Print&Scan impact. The second level increases the

information density of the QR codes, adding a way to

authenticate them.

Even if there exists several solutions for document

Tkachenko, I., Trémeau, A. and Fournel, T.

Authentication of Medicine Blister Foils: Characterization of the Rotogravure Printing Process.

DOI: 10.5220/0007565805770583

In Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2019), pages 577-583

ISBN: 978-989-758-354-4

577

and packaging authentication, the research and deve-

lopment of more efﬁcient techniques or of the exten-

sion of their application domain is a hot topic in aca-

demia and industry. While developing the novel se-

curity elements to ﬁght against packaging counterfei-

ting, the following requirements might be taken into

account: 1) easy integration and generation proces-

ses, 2) low cost, 3) use of standard printing process, 4)

fast and automatic veriﬁcation process, 5) use of com-

mon devices for veriﬁcation (ofﬁce scanners, smartp-

hones), 6) reliable veriﬁcation by any user, 7) good

level of global security and strong anti-fraud ability.

The development of copy sensitive graphical co-

des can satisfy all the items above. These codes are

based on the use of measurable but not duplicable

physical characteristics (Goldman, 1983). Today we

can ﬁnd numerous security elements that use measu-

rable but not duplicable physical characteristics and

that are sensitive to duplication attack. However, all

these elements have been developed for packaging or

documents that use the white uncoated/coated paper

and high resolution printing (from 600 dpi till 2400

dpi). Additionally, the authentication process is pro-

vided using high resolution scanners (from 600 dpi

till 4200 dpi).

We aim at developing security elements for pro-

tection of medicine packaging that are produced using

rotogravure printing on aluminum foils. To our kno-

wledge, there does not exist any tested solution for

such printing process and such type of support. The

well-understanding and characterization of the prin-

ting process is a crucial, preliminary step before deve-

loping efﬁcient security elements. Therefore, the aim

of this paper is to study the characteristics of rotogra-

vure printing process and to identify several possible

paths for authentication of such packaging.

The paper is organized as follows. We introduce

the rotogravure printing technique which is world-

wide used for medicine blister foils production in

Section 2. Then we discuss some features of the ro-

togravure process in Section 2.2 that can be useful

to ﬁght the increasing number of counterfeit medi-

cine products. We present the result of a preliminary

study of such a process, extended to foils in Section 3,

and discuss some promising outlooks for medicine

authentication in Section 4. Finally, we conclude in

Section 5.

2 ROTOGRAVURE PRINTING

CHARACTERIZATION

Rotogravure printing produces high-quality images

with intense rich colors using primary colors such as

CMYK. Each primary color is printed by one pass

through the press. Therefore, it is often used for high-

volume printing such as advertising pieces, magazi-

nes, catalogs and packaging.

2.1 Printing Process

Rotogravure is an intaglio printing process, where the

image is cut or etched below the surface. Rotogravure

printing has its pros and cons. The positive points are

long lifetime of printing cylinder (> 1 million copies),

variable printing formats, front- and backside printing

on one time, different printing primary colors, diffe-

rent materials, high printing quality, simple printing

principle. The negative points are costly cylinder en-

graving process and sloven based colors.

The rotogravure printing process uses a metal cy-

linders with an image engraved on it in the form of a

pattern with enclosed 3D cells. The cylinder radius is

determined by the ﬁnal size of the page and repeat of

the design.

The cylinders are covered by cooper as it is soft

and pliable material easy to engrave. The copper layer

is only 80 microns thick. Before engraving, the cop-

per surface is automatically polished. It removes any

imperfections and smoothed out the cylinder. Polis-

hing also adds a roughness to the surface so it can

retain the ink that will lubricate the cylinder.

An artwork needs to be created before engraving

the cylinder. Digital data for text and images feed

the engraving heads, that images the copper cylinder.

Three the more popular types of engraving process are

electromechanical engraving, chemical etching and

laser engraving.

Figure 1: The printing process using rotogravure.

The printed image results from the ink transfer from

the honeycomb shaped cells (see Fig. 1) engraved into

the copper cylinder. The un-etched areas of the cylin-

der represent the non-image (unprinted) areas. The

cell depth control the amount of ink to be transferred

to the surface. The ﬁnal size and shape of the printed

dots depend on the ink and type of paper.

After engraving, the cylinder is polished for a con-

sistent surface, and imperfections removal. The cylin-

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

578

der is electronically cleaned to remove all grease and

oil from the surface. The engraved clean cylinder is

coated with a ﬁnal layer of chromium. The chrome-

plated cylinder is polished once again to rough up the

surface, making the ink act as a lubricant. Before go

to press, the cylinder is inspected. Due to the printing

principle, one engraved cylinder is dedicated to each

primary ink.

The general scheme of printing process is illustra-

ted in Fig. 1. The cylinder rotates through an ink pan

where the cells pick up ink. The non-image areas and

over quantity of ink are scraped of the cylinder by a

blade before the ink is transferred to a paper surface.

The ink used are based on low viscosity solvents or

water. After each ink unit, the ink is dried using high

velocity air nozzle dryers.

2.2 Main Printing Characteristics

Several characteristics distinguish rotogravure prin-

ting from other well-known printing techniques as

offset or ﬂexographic.

Rotogravure can usually produce a richer black

than other processes. Rotogravure’s ability to lay

down a thick three-dimensional ink ﬁlm - altering not

only the width of the dot but also the depth - produces

brilliant colors (Keif and Goglio, 2005).

A typical characteristic of rotogravure is its ser-

rated edges on type and line work when the cylin-

der is electromechanically engraved (Keif and Goglio,

2005). Such serrated edges, not visible to the naked

eye, can be observed with a 5× magniﬁcation (Fig. 2).

(a) (b)

Figure 2: Example of serrated edges on a) the edges of a

block pattern and b) letters observed by using a ZEISS mi-

croscope with 5× magniﬁcation.

The cylinder hardness demands a smooth surface to

ensure best contact and ink transfer. The rotogravure

method is sensitive to surface defects (blade lines and

indentations). The bottom of the cell does not release

the ink quickly during the rotogravure printing pro-

cess. This phenomenon results in either ink transfer

failure so some missing dots (Fig. 3.a) or in a non-

uniform ink coverage, more precisely in ”doughnut”-

shaped dots where dots have holes in the center or on

a board of the targeted disk area (Fig. 3.b).

(a) (b)

Figure 3: Example of a) missing dots in a uniform patch

printed with resolution 152 lpi and b) ”doughnuts” in a uni-

form patch printed with resolution 136 lpi. The images were

captured using a ZEISS microscope with 5× magniﬁcation.

An additional characteristic of cylinder engraving

process is the dot coverage rate (%) that is produced

by variable depth and diameter of cells, but it always

has a ﬁxed aspect ratio. The dot coverage rate can

vary between 10% and 100%. The images in Fig. 3 il-

lustrate a dot coverage rate equal to 20% and 50%, re-

spectively. The smaller dot coverage rate tends to pro-

duce more dot imperfections (like missing dots and

closed or open ”doughnuts”) during the printing pro-

cess.

Another important element of rotogravure prin-

ting process mentioned in (Lv et al., 2010) is the ink

viscosity: the decrease of ink viscosity led to the de-

crease of the contrast of printed image, and the range

of the reachable coverage rates. The author in (Kader,

2017) showed that the low viscosity eases ink trans-

fer from the cells to the substrate and it results in a

messy print having limited sharpness. In the same

time, when the viscosity increases, the dots become

rounder, sharper and darker.

2.3 Main Production Steps

When talking about production of packaging using a

rotogravure printing process, one needs to take into

account a speciﬁc production setup. First, designing

an artwork by taking into account the printer resolu-

tion in lpi is required. The cylinder is engraved ac-

cordingly. The selected engraving method is an im-

portant factor of the quality of printed images. At the

end, the surface in contact with the engraved cylinder

(upstream inked and scraped with a squeegee) will be

more or less pressed by a roller.

All these steps are important and can inﬂuence to

the ﬁnal quality of printed image. That is why the

image resolution in artwork (in ppi), the cylinder en-

graving resolution as the printer resolution (in lpi) are

important parameters in the production setup.

Authentication of Medicine Blister Foils: Characterization of the Rotogravure Printing Process

579

3 EXPERIMENTS

In our experiments, we have used foils printed by a

rotogravure process equipped with a chemically en-

graved cylinder. The screen ruling of the cylinder was

70 lines per cm (i.e. 178 lines per inch). Blister foil

and liquid ink (foil ink) were used for the production

of samples. 780 × 880 images were captured with a

ZEISS microscope with 5× magniﬁcation (Fig. 4).

Here we imaged samples representing letter ’a’ (72

samples) and letter ’r’ (162 samples) to construct our

database.

(a) (b)

Figure 4: Several samples of our database: a,b) glyphes ’a’

with some visible enclosed ”doughnuts” (red circles), and

some ”open” doughnuts on the boundary (yellow circles),

c,d) idem but for glyphes ’r’ with more (d) or less (c) ”open”

doughnuts (yellow circles) on the vertical left side.

In these images, we remark several of the charac-

teristics of rotogravure printing as missing dots and

”doughnuts”, in particular ”open” ones (yellow cir-

cles) on a part of the border of a glyph.

3.1 Distribution of Black Dots after

Printing

After masking the acquired and registered images

with a binary mask of the ideal pattern (letter a), we

obtained the histogram depicted in Fig. 5. It shows the

distribution of gray level pixels after digital printing

and scanning. This distribution looks very approxi-

mately like a log-normal distribution (see Fig. 5).

In comparison with laser and inkjet printing proces-

ses, it is important to mention the strong asymmetry

Figure 5: The distribution of gray levels after electromecha-

nical rotogravure printing and microscope capturing pro-

cess compared with log-normal distribution. The green

curve represents here a log-normal distribution with para-

meters µ = 3.57 and σ = 0.61, hypothesis which does not

accurately models the distribution (the hypothesis was re-

jected by the χ

goodness-of-ﬁt test at a signiﬁcance level

of 0.05).

of the histogram: 74.5% of the gray levels are in the

range of 0 to 50 (Fig. 5). This reﬂects the fact that ro-

togravure printing process can produce a richer black.

3.2 Uniqueness of Printed Glyph

Due to the speciﬁc impact of rotogravure printing pro-

cess, the ﬁrst possible solution for authentication is to

use the uniqueness of printed glyph as proposed in

(Pollard et al., 2010). However, our aim is not to de-

ﬁne the speciﬁc proﬁle of the current printing, but to

show that given a character (or a glyph), each printed

pattern (e.g. printed samples representing the letter

’a’ ) has the same ”signature” as any other pattern or

”brother” printed from the same cells in the cylinder,

while its signature differs from the patterns or (”cous-

ins”) printed from cells located elsewhere on the cy-

linder. Fig. 4.a-b shows prints resulting from two dif-

ferent cylinder locations (”cousins”). To compare two

printed patterns we propose to analyze the correlation

rate that exists given a reference sample (representing

letter ’a’ then letter ’r’) with its ”brothers”, and its

”cousins” respectively.

Fig. 6 shows experimental results about 6 occur-

rences of letter ’a’ printed during three rotations of the

cylinder: two series of 18 values of the Pearson cor-

relation coefﬁcient measured with respect to a given

reference image, item 4, arbitrarly chosen in series 2.

The correlation is close to 0.62 for ”cousins” while

it stays close to 0.83 (1. resp.) for ”brothers” (items

numbered 5 and 6) in series 1 (series 2 resp.).

That results tend to show some uniqueness of the

patterns (e.g. letters) coming from a given spatial lo-

cation on the cylinder of a chemical rotogravure prin-

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

580

Figure 6: The changes of correlation values for different

captures. Samples 4, 5, 6 correspond to prints coming from

the same cylinder location meanwhile other samples corre-

spond to prints of the same letter but coming from other

cylinder locations.

ting device.

The laboratory microscope such as used is well-

suited for forensic purposes, but cannot be deployed

in practice for authentication on place on market.

Thus, we used an USB-microscope for constructing a

new database. In this database, we look at the patterns

representing a same letter (’a’ then ’r’) that were prin-

ted using one cylinder engraved by chemical process,

and one cylinder engraved by electromechanical pro-

cess respectively. Images of samples representing let-

ter ’a’ are shown in Fig. 10. As the USB-microscope

images are less resolved, features are less well captu-

red and the values of the correlation with respect to

a reference have a lower gap between the populations

of ”cousins”, and ”brothers” respectively.

In order to increase this gap, we only extracted the

edges of the letter (see example in Fig. 7).

(a) (b)

Figure 7: a) An edge of reference image sample of letter

’a’, b) an edge of another sample of letter ’a’ placed in a

different area on of the same cylinder.

The correlation with respect to the reference edge

computed with all the samples have higher values

with an increased but small gap in both chemical

(Fig. 8) and electromechanical (Fig. 9) rotogravure

printings.

Figure 8: The correlation values with respect to the edges of

a reference image for all the images in the database obtained

from ’a’-patterns and ’r’-patterns printed using a chemically

engraved cylinder.

Figure 9: The correlation values with respect to the edges of

a reference image for all the images in the database obtained

from ’a’-patterns and ’r’-patterns printed using an electro-

mechanically engraved cylinder.

We ﬁxed at both cases the authentication threshold at

T h = 0.95 which separates the population of ”cous-

ins” from that of ”brothers”. These results show the

existence of a signature of the native glyphs in the cy-

linder.

4 DISCUSSION

Exploitation of the uniqueness of printed glyphs

would mean a storage of the image captured with a

high resolution device, or a representation, in order to

allow a later comparison. In the case of the protection

of medicine foils it is not an optimal solution due to

the huge production.

4.1 Printer and Cylinder Signatures

We suppose that the authentic artwork is kept in a se-

cret. In the same time, we can suppose that each cy-

linder has its own signature when cylinders are engra-

ved using mechanical or chemical processes. Addi-

Authentication of Medicine Blister Foils: Characterization of the Rotogravure Printing Process

581

tionally, the dot shape is different for each engraving

process. Fig. 10 illustrates the differences in the shape

of letter ’a’ printed using an electronically engraved

cylinder (Fig. 10.a) and using a chemically engraved

cylinder (Fig. 10.b).

(a) (b)

Figure 10: A sample of letter ’a’ printed using a cylinder en-

graved a) electronically and b) chemically. Both cylinders

were engraved using the same artwork.

We spotted that the engraved cylinder of a speciﬁc ro-

togravure printing system may transmit its own spa-

tial ”signature” measured by image correlation. In the

same time, we can assume that the patterns printed

with the same engraved part might have higher corre-

lation values than the patterns printed using another

engraved part.

And ﬁnally, as in laser and inkjet printers, each

printed character contains the printer signature. The

only thing we need to study is the impact of each sig-

nature. That means we need to test either the cha-

racters differ more from printer to printer (we ﬁx the

cylinder and we change the printer used) or the cha-

racters differ more from cylinder to cylinder (we ﬁx

the printer and we use some cylinders engraved using

the same method).

Table 1: Different combinations of cylinder and printer in

order to understand whose signature is more important du-

ring authentication.

Cylinder Printer

Fixed Variable

Variable Fixed

Variable Variable

In order to understand which signature is more im-

portant, we need to analyze different combinations of

printer and cylinder that are presented in Table 1.

4.2 Authentication System based on

Cylinder Signature

However when we talk about the counterfeiting of

packages that were printed using rotogravure process,

we need to take into account the necessity to produce

new engraved cylinder as a counterfeiter do not have

access to the original cylinder. That is why the printer

forensic investigation or more precisely the cylinder

forensics can be a good alternative for detection of

counterfeits.

Let W be the original artwork image after cylinder

engraving and printing process this image will be I

W+N

, where N

is a noise added by the rotogravure

printing. We suppose, that all images I

that were

printed using authentic cylinder C

have some speciﬁc

characteristics added by the cylinder signature. That

is why, we can tell that all these images belong to one

class C

(∀I

∈ C

In the same time, even if the counterfeiter produce

the same artwork W for production of counterfeiter

cylinder C

, the images printed using this cylinder will

be different I

= W + N

∈ C

due to the signature of

counterfeiter cylinder C

Thus, the authentication test for a new captured

image I

can be formulated as a hypothesis test:

: I

∈ C

: I

/∈ C

where the image I

is authentic when the hypothesis

is accepted, otherwise the image I

cannot be con-

sidered as authentic.

4.3 Possible Attacks

In this section, we want to list the possible counterfei-

ter strategies while faking the blister foils.

1. Photo-printing (scan-and-engraving) attack. This

attack consists on scanning the blister foil of au-

thentic medicine and using this scanning image as

an artwork for engraving a cylinder. This attack is

not realistic, as the image after scanning is noisy

and cannot be used for cylinder engraving without

pre-pocessing step.

2. Scan-Processing-Engraving attack. Here an op-

ponent tries to estimate the original artwork (that

was used for authentic cylinder engraving) using

the scanned image.

3. Artwork counterfeiting-and-engraving attack. In

this case, an opponent creates his/her own art-

work, that must be close to authentic artwork. If

the authentic artwork is simple, the recreation pro-

cess can be very simple. That is why the smart

construction of authentic artwork with non-trivial

security elements is a very important process.

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

582

5 CONCLUSIONS

The protection of medicine blister foils is an impor-

tant issue these days. There exists a big amount of

security elements that were developed for document

and packaging protection, and anti-copy authentica-

tion. However, all the operational solutions have been

developed for laser and inkjet printers and used with

the uncoated/coated paper substrate.

In this paper, we are focusing on rotogravure prin-

ting techniques and blister foil substrates that are used

for production of medicine packaging. We have lis-

ted the main characteristics of these production pro-

cess and we have presented the preliminary study of

this process. We have done the experiments in order

to show the distribution of black color after printing

and capturing process. We have shown the unique-

ness of patterns printed by rotogravure. Finally, we

have discussed the future path where we want to un-

derstand the impact of cylinder and printer signature

in this uniqueness. Based on this study, we will con-

struct a new security element for rotogravure printing

on blister foils.

ACKNOWLEDGEMENTS

This work was funded by project PackMark supported

by the Indo-French Center for the Promotion of Ad-

vanced Research (IFCPAR) under contract IFCPAR-

7127. All the printed samples were provided by Ser-

gusa Solutions Pvt Ltd.

REFERENCES

Goldman, R. N. (1983). Non-counterfeitable document sy-

stem. US Patent 4,423,415.

Kader, M. E. A. (2017). The impact of ink viscosity on

the enhancement of rotogravure optical print quality.

International Design Journal.

Kee, E. and Farid, H. (2008). Printer proﬁling for forensics

and ballistics. In ACM workshop on Multimedia and

security, pages 3–10. ACM.

Keif, M. G. and Goglio, T. (2005). Identifying high-volume

printing processes. Visual Communications Journal,

pages 35–42.

Lv, X., Liu, C., Wu, Y., and Ipsen, H. (2010). Variation of

gravure printing characteristic curves. In 17th IAPRI

World Conference on Packaging.

Navarro, L. C., Navarro, A. K., Rocha, A., and Dahab,

R. (2018). Connecting the dots: Toward accountable

machine-learning printer attribution methods. Journal

of Visual Communication and Image Representation,

53:257–272.

Picard, J. (2004). Digital authentication with copy-detection

patterns. In Electronic Imaging 2004, pages 176–183.

International Society for Optics and Photonics.

Pollard, S. B., Simske, S. J., and Adams, G. B. (2010). Mo-

del based print signature proﬁle extraction for forensic

analysis of individual text glyphs. In Information Fo-

rensics and Security (WIFS), 2010 IEEE International

Workshop on, pages 1–6. IEEE.

Tkachenko, I., Puech, W., Destruel, C., Strauss, O., Gaudin,

J.-M., and Guichard, C. (2016). Two-level QR code

for private message sharing and document authentica-

tion. IEEE Transactions on Information Forensics and

Security, 11(3):571–583.

Authentication of Medicine Blister Foils: Characterization of the Rotogravure Printing Process

583