The first part of this series discussed letterpress numbering machines, detection of counterfeit serial numbers by printing process examination, and strategies for optimizing letterpress print characteristics. The second part covered security inks, visible color, ultraviolet and infrared responses, multicolor serialization, and metallic effects.  Printing process and security inks are just two components of serial number security and are supplemented by a third element: font design.  Because most serial numbers are small and the limited art can be replicated relatively easily compared to intaglio or offset artwork, the importance of font design may not be apparent.  But unlike static security artwork, recognizing that serial numbers are variable data is critical to understanding why serial number fonts can and should be optimized.  This paper begins a larger case for font design innovation by reviewing some basics and contemporary strategies. 

All strategies discussed in this paper are for informational purposes only.  Document manufacturers and issuers must decide which approaches, if any, are appropriate for the needs of a specific document or are compatible with their own manufacturing workflows and quality control practices.  Although the focus of this series is letterpress serialization, some concepts might be adapted for digital serialization technologies and will be addressed in greater detail later in this series. 

Font design vs. printing process identification

Figure 1 was first presented in Part 1 of this series and displays three counterfeit serial numbers detectable with magnification because they lack typical letterpress numbering machine print microscopic characteristics, which include a halo, voids and/or debossing.  Because letterpress print characteristics are generally expected paper security document serial numbers, a serial number that does not exhibit letterpress print characteristics may be regarded as suspicious even if the document is unfamiliar and the expected serial number format is unknown. 

Figure 1. These three counterfeit serial numbers are suspicious because they lack microscopic     characteristics associated with letterpress printing, especially a halo.  This kind of counterfeit detection based on print characteristics requires magnification but can be made without prior knowledge of the expected security features of the document in question. Compare to Figures 2 and 3. 

However, not all counterfeiters use incorrect serialization printing methods, and some do have letterpress capabilities.  In Figure 2, the counterfeit serial number has the expected letterpress halo and debossing, making it hard to identify as suspicious based on printing process characteristics. 

Figure 2. This counterfeit serial number shows a letterpress halo with some debossing, making it difficult to detect solely by printing process.  Not all counterfeiters can or do use letterpress to serialize counterfeits.  Nonetheless, this example shows why letterpress should be combined with security ink and a custom font, so security does not rely just on a printing process. Compare to Figures 1 and 3. 

Figure 2 reinforces that letterpress print characteristics are just one facet of serial number security and should be supplemented by a security ink and a customized font.  The role of ink was covered in Part 2 of this series, but here Figure 3 illustrates detection of counterfeit serial numbers by font design.

Figure 3. Reflected light images of the same four counterfeits in Figures 1 and 2 (bottom row), next to letterpress numerals from genuine documents of the same types (top row).  Consider which counterfeits can be identified effectively by printing process characteristics, or differences in font design, or both.  Font comparison can be done without magnification, but only if the genuine font design is known. 

Instead of identifying printing process characteristics as in Figure 1, for Figure 3 compare the shapes of genuine numerals in the top row to counterfeit numerals in the bottom row, looking for graphical differences more than print characteristics.  Counterfeit detection by printing process requires magnification but not prior knowledge of the expected font design, while detection of an incorrect font requires that the genuine font be available for comparison, but magnification may not be needed. 

Font customization

As Figure 3 shows, a custom font can contribute to security even if serial number graphics are relatively simple because counterfeiters can still introduce errors.  Further, font design is inexpensive with limited downside risk, and the font need not stand alone if it is paired with letterpress printing (from Part 1) and security ink (from Part 2).  Accordingly, it is common for serial number fonts to be customized and Figures 4 through 7 illustrate some examples in issued security documents. 

For counterfeit resistance, some common ways of customizing numerals are shown in Figure 4 and include extraneous graphics, added or exaggerated serif size or placement, and exotic character shapes.  None of these prevent counterfeiters from capturing and replicating font artwork, but any redrawing or editing work forced on counterfeiters provides opportunity for error. 

Figure 4. Serial number fonts in some issued security documents, showing graphical stylizing designed to resist counterfeiting, alteration, or both.  A custom font helps with counterfeit resistance, but alteration resistance is improved by exaggerating differences between characters that sometimes look alike (e.g., 3 and 8).  In most but not all of these, the exact same character set is used in every position.

For alteration resistance, compare numerals within the serial numbers shown in Figure 4, particularly those that sometimes look like one another (such as 1 and 7, 3 and 8, 8 and 0, 7 and 9, and so on).  In many of the fonts in Figure 4, graphical differences between these numerals have been exaggerated so alteration attacks require more extensive editing. 

Figure 5 shows some serial numbers that include a custom character.  Counterfeiters can simulate custom character artwork just as they can simulate numbers and letters, but a custom character might convey issuer branding, identify a document version/series, contribute to check digit calculations, or perform other functions that may not be related exclusively to security. 

Figure 5. Serial number font designs that include custom prefix or suffix characters.  Custom characters or prefixes/suffixes may communicate issuer branding, identify a document series, support check digits, or play other roles.  Inclusion of custom characters in a serial number may stop counterfeiters from buying a commercial numbering block with the same character, but do not impede counterfeiting much. 

In the machine fonts in Figure 6, the exaggerated differences in numeral shape help machine readers differentiate between numerals more reliably, though alteration resistance could be an added benefit.  Like alteration resistance, machine readability depends on more than font design and can be facilitated by serial number inks with specific ultraviolet, infrared, magnetic or other properties. 

Figure 6. Letterpress font and barcode designs are likely used for machine reader applications, including numerals designed with shapes intended to look very different from one another.  This approach can also improve resistance to alteration because it requires more extensive editing to make one numeral look like a different numeral, which may make the alteration more difficult to conceal. 

Figure 7 is repeated from Part 1 of this series and illustrates how relief plates offer two design platforms: first, the contour of each numeral that defines the macro font design, and second, textured artwork on the relief surface that forms a pattern inside the image area of the font.  In Part 1, this concept was discussed as a mechanical characteristic unique to relief printing methods.  Here it is presented again to state that the font and the texture are separate design canvases, and both can be optimized.  Design strategies for relief plate surfaces will be revisited in a later part of this series. 

Figure 7. From Part 1 of this series, a letterpress serial number impression in which the two left numerals are printed from textured relief surfaces and the two right numerals from smooth relief surfaces.  Texture in a relief surface is another canvas upon which artwork can be applied and might be considered both complementary to and an extension of the macro font design.

Expanding the character set

Applying the same serial number to multiple counterfeits increases detection risk, so counterfeiters may collect a library of individual serial numeral graphics from multiple genuine documents of the same type.  Once assembled, the complete numeral set can be transposed between counterfeits to generate new serial number combinations.  If every wheel in a numbering machine contains an identical set of (more or less) ten numeral graphics, any serial number applied by that machine is merely a reordering of the same ten artwork elements.  Such a small character set does not encompass much graphical complexity and the labor required to capture, edit, and transpose numerals is minimal.  Unfortunately, most of the examples shown in Figures 4 through 6 contain only ten numeral graphics.  Can this be expanded?

A simple way to grow the total character set in a security document with two serial numbers is to use a different font for each number.  Figure 8 shows left/right serial number pairs in banknotes, for which the second font effectively doubles the character set from 10 to 20 numerals.  This expands the quantity of numeral graphics a counterfeiter must manage, increases counterfeiter labor, provides more opportunity for counterfeiter error and font design costs almost nothing for genuine document issuers.  The extra numerals modestly increase editing labor for counterfeiters, but the security benefit is minimal because the number of additional numerals is also minimal. 

Figure 8. Left/right serial number pairs in banknotes, printed in different fonts but in the same or similar ink color.  The size and font shape of a numeral seems to be identical within each string, regardless of position.  Use of two fonts does not provide a major improvement in security.  However, it does double the total character set and lays the groundwork for further ideas about character set size.   

Theoretically, if expanding the character set from 10 to 20 numerals adds modestly to counterfeiting labor, can expanding it from 20 to 70 or even more numeral graphics provide a larger increase in counterfeiting labor and workflow complexity?  As is often the case with design strategies, the answer is a qualified yes.  Novel numbering can be one way to achieve this, in certain implementations. 

Novel numbering

“Novel numbering” is a term used to describe letterpress serialization that shows a visible size progression, like the examples in Figure 9.  Novel numbering forces counterfeiters to do additional editing work if they capture a numeral graphic from one position from a genuine document and transpose it between different positions in multiple counterfeits, since the size of the graphic must be adjusted to match its position in each unique string.  But novel numbering can be more than just size progression and may also expand the total serial number character set, in certain implementations.  

Figure 9. Horizontal and vertical novel numbering in letterpress serial numbers, in which the font size and/or baseline position of a numeral depends on its position in the string.  Custom novel numbering machines prevent counterfeiters from purchasing an identical commercial numbering machine, but do not stop digital counterfeiters from capturing, resizing and otherwise manipulating the font design. 

Beginning with a typical novel numbering implementation in which numerals are proportionally scaled both vertically and horizontally between positions, the left side of Figure 10 shows examples of two serial numbers in which font size gradually increases.  The right side of Figure 10 shows height-normalized images of the same numerals captured from different positions in the string, which illustrates whether each position has a different numeral design or the same numeral shape at a different font size. 

Figure 10. On the left are two serial numbers with novel numbering and repeated numerals in various positions.  At the right of the figure, compare the height-normalized images of small and large numerals.  Despite the size change, the shapes of the small and large numerals look the same, showing that the change in numeral size does not also mean a change in font/shape. Compare to Figures 11 and 12 (below).

In Figure 10, scaling of the font was done proportionally in the horizontal and vertical dimensions, so large and small numerals are otherwise of the same shape.  Setting aside the size differences and considering only shape, the full numbering machine used to apply Figure 10 likely includes only ten different numeral shapes.  Counterfeiters need only capture or redraw each of the ten numerals once to complete the character set and then resize them as needed to transpose between positions, so this novel numbering approach adds labor but not much complexity to counterfeiting workflows. 

Numeral distortion

Following on Figure 10, a different implementation of novel numbering is shown in Figure 11.  Like Figure 10, Figure 11 shows a novel numbering string in which characters cannot simply be reordered without enlarging or shrinking the font to the correct size.  But when the numerals in Figure 11 are height-normalized just as in Figure 10, it shows larger numerals are not proportionally scaled both horizontally and vertically as in Figure 10 but are stretched versions of shorter numerals.  Stated differently, in Figure 11 the numeral “3” in the third position shows less distortion and the numeral “3” in the seventh position shows more distortion because they are roughly the same width but are of different heights. 

Figure 11. On the left is a serial number featuring novel numbering and repeated numerals.  At the right of the figure, compare the height-normalized images of short and tall numerals.  In this string each numeral occupies about the same horizontal space, but their heights vary. The taller numerals are not the same shape as the shorter numerals since they are stretched more. Compare to Figures 10 and 12.

Each wheel in Figure 11 appears to have its own unique character set because the degree of distortion in any numeral depends on its position in the string.  Assuming each of the seven wheels in the numbering machine used to apply the serial number in Figure 11 contains ten unique numeral shapes (0-9), then the total character set is 7 X 10 = 70 unique numeral shapes, a significant expansion over the ten numeral shapes anticipated in Figure 10.  The distorted numeral shapes in Figure 11 are arguably more complex than the proportional scaling in Figure 10 since they produce not only a size change but also a shape change.  However, neither of these approaches is very complicated. 

Rotation and baseline position

Prior examples included position-dependent scaling based only on size (Figure 10) or both size and distortion (Figure 11).  Rotation and/or baseline position are some additional variables that can be position-dependent in a serial number string, as shown in Figure 12.  The full serial number in Figure 12 appears curved because the horizontal placement, rotation and font size of each numeral depends on its position.  Each “0” has a different appearance, suggesting each wheel in the numbering machine may feature its own custom rotated and resized character set. 

Figure 12. High and low magnification images of a serial number in which horizontal placement, rotation and/or size depend on a numeral’s placement within the string.  Customizing the string globally is different than customizing individual numerals at the font level.  This example is oriented vertically, but the same concepts can be applied to horizontal serial numbers. Compare to Figures 10 and 11. 

In theory a larger character set makes counterfeiter numeral transposition more complicated because there are more numeral designs to manage, and more work required to assemble the full collection.  The scaling in Figure 10, distortion in Figure 11, and rotation/placement in Figure 12 demonstrate valuable concepts.  But Figures 10-12 only incorporate simple manipulations that counterfeiters can recognize quickly and reverse engineer relatively easily.  Accordingly, later parts of this series will explore ways to make serial number font designs more challenging for counterfeiters to reverse engineer by designing every numeral as a natively unique graphic, not just a derivative of or variation on a base font. 


The above contemporary strategies for serial number artwork complement earlier parts of this series about letterpress and ink optimization, which together form three critical components in serial number design.  Serial number artwork is small and appears to offer only limited complexity, but in truth variable serial numbers provide opportunities to disrupt counterfeiter workflows that do not exist in the context of static intaglio or offset artwork counterfeiting.  This is because one printed serial number contains only a small subset of the characters present throughout the wheels of the numbering machine, the numeral graphics can be related to their position within a serial number string, and counterfeiters are motivated to transpose serial number graphics to ensure multiple counterfeits do not carry identical serial numbers.  Subsequent work in this series will expand on these font concepts, with the goal being development of novel strategies that apply letterpress serial numbers in unexpected ways.

Disclaimer: This document represents the opinions of its authors and not necessarily the opinions of the U.S. government.  The technologies and strategies described may not be available, appropriate, or manufacturable for all document issuers.  The examples shown do not imply anything about the quality of a document, its designer, its manufacturer, or the issuing authority. For informational purposes only. 


Join the conversation.

Keesing Technologies

Keesing Platform forms part of Keesing Technologies
The global market leader in banknote and ID document verification

+ posts

Joel Zlotnick is employed by the US Department of State, Bureau of Consular Affairs, Counterfeit Deterrence Laboratory as a physical scientist. His current work involves research in security artwork and design techniques in security printing. He is an instructor on counterfeit detection at the US Department of State Foreign Service Institute.

+ posts

Tyra McConnell is a Forensic Document Examiner at the US Department of State, Bureau of Consular Affairs, Counterfeit Deterrence Laboratory. She provides training on security documents and develops presentations and e-learning courses regarding counterfeit detection.

Previous articleMozambique Launches a New Family of Banknotes
Next articleJapan’s New Banknote Series