Obfuscation of Embedded Codes in Additive Manufactured Components for Product Authentication

Abstract

Enhancement in the capabilities of additive manufacturing (AM) methods has led to development of many high-value components for aerospace, automotive and medical fields. Security concerns, such as (a) a predominantly cloud based process chain of AM may be breached and stolen files can be used for unauthorized reproduction of parts and (b) legitimately acquired parts can be reverse engineered, need to be addressed for this field to protect intellectual property and deter counterfeiting or unauthorized production. In the present work, a method of embedding an identification code inside the parts manufactured by AM methods is presented, which takes advantage of the layer-by-layer manufacturing process. The code is obfuscated by segmenting it into a specific number of parts, which are distributed throughout a large number of printed layers. In this case, viewing the code only from a specific direction is provides the correct visualization. A further obfuscation scheme is demonstrated that embeds multiple identification codes in the interpenetrating format. Only a specific set of processing conditions can lead to printing of the authentic code inside the part correctly. Numerous other conditions lead to printing of wrong code inside the part, which will lead to positive identification of counterfeit or unauthorized parts. Securing the AM process chain can help in accelerating the industrial applications of this versatile method.

Graphical Abstract

A method to embed and obfuscate an identification code in computer aided design models intended for additive manufacturing is presented to work against counterfeiting and reverse engineering. Two interpenetrating codes are segmented into small parts and embedded into numerous layers. The default slicing and printing options retain the code that reads “counterfeit”. A predetermined specific set of processing conditions retains the authentic code.

1. Introduction

Additive manufacturing (AM) methods are now moving from prototyping to mainstream production methods. Security of product design and intellectual property (IP) is an important issue for industries as diverse as aerospace, automotive, and medical.^[1–5] Recent advances in AM methods have brought new focus on security for two reasons (a) extensive use of computers and cloud based resources in the entire process chain, which makes it vulnerable to cybersecurity threats and (b) wide availability of high quality three-dimensional (3D) printers that can facilitate production of high quality parts by unauthorized personnel.^[6–8] In traditional manufacturing, such as machining and casting, the skill of the operator is an important component in obtaining high quality parts with correct dimensional tolerances. In addition, the infrastructure required for those traditional methods is not as widely available and cost effective as 3D scanners and printers. Therefore, unauthorized production of parts from stolen files and reverse engineering are significantly greater concerns in the AM process chain compared to most traditional manufacturing methods.

The cybersecurity scenario in AM has been rapidly evolving. While new methods for product authentication and quality assessment are being developed specifically for parts produced by AM techniques, unauthorized production and reverse engineering still remain major challenges. AM is being used for high value components such as aircraft engine parts, one of a kind spacecraft parts and medical devices.^{[1, 2, 9, 10]} In these industries, cyberattacks can lead to safety concerns from use of AM parts due to the possibility of sabotage, compromised IP and substandard raw materials.^[11–15] Therefore, methods are required to positively identify parts manufactured by a genuine AM process chain.

A variety of cyberattacks are possible in industrial AM scenario. For example, many 3D printers are connected to Internet for remote operation, monitoring and diagnostics. Vulnerabilities in these connected printers have been found,^[16] where software or hardware can be hacked to overheat the printer and cause fire.^{[17, 18]} Similar hacking incidents may target the model geometry and introduce defects or design mutations during the printing process. The layer by layer manufacturing process of AM may take several hours, even days, to print a large scale part containing micro-sized details. In such cases, small internal defects introduced on the go by hackers may be difficult to characterize. Many of these external threats can be addressed by strengthening the network level security, e.g., by controlling the network access, installing firewall, creating a manufacturing subnet and using stronger passwords and file encryptions. However, internal malicious actors may cause similar attacks and bypass all these security measures. It is a growing concern that these scenarios need to be addressed by developing new security approaches rooted in the product design to help the AM field advance rapidly for adoption in industry for creating parts that can be used in service.

The present work is focused on the specific scenario where authentic AM parts need to be identified. This scenario addresses the possibilities such as unauthorized production and reverse engineering from a genuinely acquired part. Existing methods such as stamping of serial number or surface labeling are being used for AM parts as well.^[19] However, reverse engineering of the same part using AM can produce clones of a product with the same serial number and makes it difficult to identify the original part. The present work takes advantage of layer-by-layer manufacturing in AM process to embed an identification code in the part. Schemes are presented to obfuscate the code in the CAD models as well as in printed part. First, the code is sliced into a large number of segments and converted to a cloud of voxels to make it difficult to copy it by reverse engineering. These voxels are embedded into numerous layers of the sliced part. Second, in order to further obfuscate the code inside the CAD file and reduce the possibility of copying or changing the code geometry in the stolen CAD files, two different interpenetrating codes are embedded. The default or obvious options of slicing and printing lead to survival of the code that reads “counterfeit”. However, a pre-determined specific set of processing parameters, such as slicing and printing orientation, provides the correct code in the printed part. These embedded codes can be recovered from imaging methods such as micro-computer tomography (CT) scan or x-ray radiography. This kind of security scheme can be used in addition to other possible network or file level security measures and can help in identification of genuine parts if the password or file encryptions are compromised successfully to print counterfeits.

In the following subsections, an overview of some of the anti-counterfeiting methods that have been developed for the AM field is presented, followed by a description of the scheme to embed the obfuscated code in an AM model. The core principal of the presented obfuscation scheme is independent of the the type of materials and it expected to work for polymer, ceramics, and metallic materials. In addition, the code obfuscation parameters should be adapted to the 3D printing technology used for the production of the part because each technology has a different resolution, minimum printable feature size and layer thickness. Here, material extrusion-based fused filament fabrication (FFF) and material jetting based PolyJet technologies are used to demonstrate the proposed obfuscation security methods. Use of transparent materials has allowed visualizing an embedded code of a colored material. However, micro-CT scan is performed to obtain the images of the embedded code, a process that is expected to be similar for opaque materials such as other polymers and metals.

1.1. State of the art in AM product authentication tools

AM includes a variety of methods such as FFF, laser sintering, laser melting, and ultraviolet curing, among others.^[10] This variety of fundamental principles used as manufacturing methods makes it challenging to develop security and product authentication tools that apply across the entire spectrum of AM methods, which reflects in the available security and product identification methods currently used in this sector. A number of anti-counterfeiting marking methods and labels are available for various purposes.^[19–21] For example, ultraviolet (UV) security marker pens; watermarking; tamper-evident label seals, security tapes and other packaging products; and holographic labels are used on hardware parts. These same marking methods can also be used on AM parts. However, due to surface application, these markings are easy to temper with and may wear off in shipping and handling or use of the part. The modern authentication and security tools use an automatic identification and data capture (AIDC) process to identify and track items in a wide range of applications including supply chain management. The AIDC family covers a broad range of technologies including barcodes, radio frequency identification (RFID), biometrics and integrated circuit cards among others.^[22–24] These methods can also be applied to AM process chain and products for tracking and authentication. For example, RFID tags have been integrated in the AM products.^[25] Many of these technologies are suitable only for a select few low-temperature AM methods due to the possible damage at high temperatures encountered in many common AM methods including FFF and slective laser sintering or melting.

Interest is rapidly increasing in the area of integrating sensors in AM parts.^{[26, 27]} These sensors are integrated with the aim of structural health monitoring by means of measuring temperature, strain or stress.^[28–30] Some of these integrated sensors can be used for the additional purpose of product authentication and anti-counterfeiting. In addition, a master audio fingerprint can be generated for each 3D printing process to verify the part’s integrity when printed again.^[31] However, such methods are not yet fully developed. Various authentication and anti-counterfeiting technologies are at different stages of development and commercialization. A list of existing commercialized tools for the AM sector is summarized in Table 1.^[32–47] It can be observed that most of the solutions are implemented at the network or software level and include solutions such as encryption and access control. Most of these technologies are also extensions from traditional manufacturing methods and process chains, which are extended to AM without taking advantage of functioning of the actual AM methods.

Table 1.

Summary of a list of existing commercially available authentication and security tools.^[46]

Company	Security technology
Identify3D	Identify3D software tools are designed to secure the 3D design-to-manufacturing process end-to-end. The “Protect” software encrypts designs to allow only a fixed number of builds on a particular hardware with designated expiration date; the “Manage” software controls the encrypted designs’ distribution and authorization; and the “Enforce” package provides verification, decryption and build instructions to the authorized end user [32].
Treatstock	This online digital blueprint-sharing platform is developing a system where design owners can integrate uploaded standard tessellation language (STL) file with a hidden “watermark” for tracking and verification for authenticity [33].
Authentise	Their 3DIAX all-in-one additive manufacturing developer platform includes cloud storage facility, rendering, parameter control, model manipulation, slicing and computer vision-based print monitoring tool [34].
CGTrader	An online marketplace where the user can download a non-printable version of the design to review and validate without having to pay for the fully printable version. This two-step process can help reduce the cost of content creation and minimize return of printed parts [35].
Kabuni	Kabuni builds the blockchain platforms that protect, print and pay for 3D printing at industrial scale to secure the digital thread and prevent unauthorized 3D printing [36].
LINK3D	LINK3D integrates blockchain technology for industrial 3D printing to preserve file integrity and traceability, and pre-verify orders to facilitate matching and authentication [37].
Disney	Use of anti-scanning materials on surfaces of objects, which are light absorbing or reflect light in unconventional directions, to deter reverse engineering [38, 39].
Rize Inc	Rize introduced a system for ink jetting voxel-level QR codes on 3D printed parts, so that the user can visualize and use a smartphone app to scan the QR code and authenticate the part [40].
Karlsruhe Institute of Technology (KIT) and Zeiss	Embedding tiny fluorescent microstructures to foil 3D printed counterfeits. These microscopic features are about 100 μm in length for reading by special readout instruments [41].
Nanoscribe Germany	Developed the process of fabricating multi-level “diffractive optical elements” (DOEs) in a single step, which is faster than conventional planar lithography methods. DOEs generate light distributions arbitrarily in the far field for beam shaping and can serve as security feature [42].
Sofmat UK	An anti-counterfeiting 3D barcode marker with a 0.4 μm feature size that is 3D printable directly on objects so that they can be tracked and verified using a laser scanner [43, 44].
InfraTrac	InfraTrac adds taggants by directed energy deposition 3D printing technology to parts as a chemical fingerprint that can be read by x-ray. The invisible taggants 3D printed on surfaces of parts can be detected by x-ray fluorescence on titanium samples [45].
Quantum Materials Corporation	This technology integrates quantum dot with AM for securing 3D printing against anti-counterfeiting. The embedded quantum dots serve as unique, physically uncloneable signature known only to the manufacturer [46].
Applied DNA Sciences, Inc.	Botanical deoxyribonucleic acid (DNA) taggants to mark parts as genuine for supply chain security in the electronics industry [47].
3DP Security, Inc.	Features to embed in CAD models to make visualization different from the actual printed part geometry.

An approach of embedding quick response (QR) codes in AM parts was recently demonstrated, which is illustrated in Figure 1.^{[48, 49]} In this approach a QR code was segmented into multiple small segments which were converted to a cloud of segments in such a way that these segments would be distributed over a large number of layers in the sliced model. The CAD model containing the cloud of QR code segments is converted to STL format, which is then sliced to capture the cloud of segments in a large number of slices. This model is converted to G-code and then printed using a 3D printer, where these QR code segments are either printed with a secondary support material or as voids depending on the actual 3D printing technology used. The embedded code is read using a micro-CT scanner and the acquired image is processed to enhance the contrast and read the code. The obfuscation scheme of slicing the code into a large number of segments provides advantage that the code can be read properly only from one specific viewing angle or from 180° to that direction. Any misalignment from this perfect direction makes the code unable to scan. Each segment of the code can be printed in a different layer so that the effect of such codes on mechanical properties can be minimized. Although this method of obfuscation is effective, it is not very strong. The present work shows a new method of obfuscating the code and using two interpenetrating exploded codes, which provides much greater security and significantly reduces the reverse engineering risk.

Figure 1.

Workflow in additive manufacturing from CAD solid model development to the final physical model printing stage.

1.2. Proposed AM product security approach

The method presented in this work relies upon embedding a set of two interpenetrating QR codes in the CAD model. Default options of slicing and G-code conversion result in preserving the code that reads “counterfeit”, while processing of files under a pre-determined specific set of parameters results in preserving the correct QR code. An imaging method such as radiography or tomography can be used to read the embedded code to determine the authenticity of the part.^[48] Apart from selectively printing the QR codes, the same scheme can be used in many other beneficial ways. For example, in one case, this scheme can lead to disappearance of all codes from the product to provide high quality fully dense part, while the incorrect processing conditions or settings will lead to printing of a large number of segments. Depending on the printing method, these segments can be made of support materials or voids, which will reduce the mechanical properties. The combination of design and printing conditions specified for this part helps in creating a security method. Since the printed part will have only one QR code, the reverse engineering of the product will provide a CAD model that will be different from the original file. It is also anticipated that the need of reverse engineering hundreds of small features in the part to create the QR code will be a deterrent against counterfeiting.

2. Materials and Methods

SolidWorks^® 2016 (Dassault Systemes, France) 3D modelling software is used for CAD modeling. All STL files for the corresponding CAD model are generated in SolidWorks using pre-determined settings (Coarse Resolution). All toolpaths shown in this work are generated by importing STLs and slicing using CatalystEX (Stratasys, USA) software under the following slicing properties : 0.1778 mm layer resolution, solid model interior, SMART support fill, and milimeters STL units. CatalystEX is the designated slicer software for the Dimension Elite™ FDM 3D printer (Stratasys, USA), which is used for preparing toolpath files for 3D printing the grey color thermoplastic specimens (ABSplus-grey filament, Stratasys, USA). Any solid geometry will be 3D printed by the ABSplus-grey model materials, while any internal porosities or support structures will be 3D printed by the support materials (Soluble Support Technology, Stratasys, USA). The transparent photopolymer specimens are sliced and processed by the GrabCAD Print software and then 3D printed using the J750 Color Printer (Stratasys, USA) (pre-calibrated layer resolution in 0.027 mm). The 3D printed parts are scanned using a Bruker SkyScan 1172 X-ray micro-computed tomography (micro-CT) scanner (Micro Photonics Inc, USA). The micro-CT scan models reconstruction is conducted using NRecon software (Micro Photonics Inc, USA), processed in the CT-Analyzer (CTan^®) analysis software (Micro Photonics Inc, USA), and exported as an STL file that is opened and rendered as graphics body in SolidWorks^® 2016.

3. Design of obfuscated QR code

3.1. Obfuscating QR code by segmentation

The QR code (25.55×25.55 mm²) is divided into 191 rectangular segments (width 1.0 mm, thickness 0.75 mm, and lengths varying from 1.0 to 7.2 mm). Although significantly smaller codes can be embedded in parts, this size is selected for possible visual inspection of the printed code in the part. Some of the material jetting and metal printing methods can print features of 15–50 μm size, which would make it possible to print codes of smaller than 1×1 mm². A QR code image is imported as a sketch image in SolidWorks, then the Sketch function is used to generate 2D rectangles based on the segmentation scheme mentioned above. These QR code segment sketches are then extruded using the Extruded Surface function in SolidWorks to generate open cuboids embedded inside a solid spherical (diameter 40 mm) enclosure in a CAD model as shown in Figure 2. The spherical shape of the model is selected for visualization purposes. This shape can be rotated in any direction to visualize the embedded code segments. In the example, the desired QR code leads to the webpage “http://engineering.nyu.edu/composites/”. Following the AM workflow described in Figure 1, this model is tessellated to STL and sliced and then the toolpath files are generated. The slicing results depend on a number of parameters such as thickness of each segment, spacing between segments and slicing layer thickness. If each segment of the code is thinner than the slicing layer thickness, it can be completely lost during slicing. However, selective placement of various QR code segments in the CAD model helps in preserving it through the slicing process if each segment thickness is greater than the slicing layer thickness. Figure 2 shows an example of the possibility of selectively printing the QR code, where the same STL model is placed under different slicing orientations. The QR code does not appear in the solid spherical model under default slicing orientation. In addition, it is observed that the 90º rotation with respect to X- or Z-axis for slicing also does not result in any QR code in the sliced model. Only 90º rotation with respect to the Y-axis successfully retains the embedded code. The embedded QR code either disappears or is retained only partially at other slicing angles. Additional testing is conducted to determine the efficacy of the rotation scheme by rotating the cube at 0.1° in each step. It was found that the code survives for rotation up to 98°. Increasing the rotation further with respect to the Y-axis does not result in survival of the code in the model. This result shown in the material extrusion-based FDM polymer 3D printing process is dependent on the thickness and slicing resolution of the code’s CAD model and can be tuned by these parameters.

Figure 2.

CatalystEX (Stratasys, USA) slicing results of the CAD model (SolidWorks^® 2016) with an embedded obfuscating QR code, out of 4 slicing orientations, only the Y-90º (rotate 90º about Y-axis from the default orientation) results in printing of the embedded QR code.

For obfuscation, the code can be designed to appear at any pre-determined slicing angle and direction, making it difficult to determine the actual orientation and settings needed to print the code in the part. Since the embedded codes can be designed at any angle to the three principal planes, determining the exact angle for slicing can be very challenging even if it is tried visually using the CAD file. Especially, the time and effort required to decode the precise orientation may be a deterrent for the attackers. It is also anticipated that obtaining the exact code in CT-scanned parts for reverse engineering will be very time and effort consuming because of small feature size of each segment of the code and a large number of segments present in the part.

In the present work, the QR code is segmented into 191 parts of 25 vertical separations and embedded in 23 different layers, where the layers are separated by 1.25 mm each. This segmentation scheme is used for validation of relationship between feature dimension and printer resolution. For example, the CAD model shown in Figure 2 has minimum feature dimensions of 1.0×1.0×0.75 mm³, which can be sliced to retain all segments because the slicing and printing resolution is 0.1778 mm, around 4 times smaller than the feature thickness of 0.75 mm. The segmentation scheme and design methodology can be used to preserve the QR code in the 3D printed part for product authentication and verification.

3.2. Obfuscation by two interpenetrating codes

Two interpenetrating QR codes are segmented in the way described above and embedded such that processing the files using default settings will print a code that reads as “counterfeit”, whereas a carefully controlled process flow will print the genuine code inside the part.

3.2.1. Case study 1: printing with material extrusion technology

In this case, two QR codes are segmented into a total of 354 rectangular segments and positioned at certain pre-determined angles in the CAD model as shown in Figure 3. The authentic QR code (18×18 mm²) is divided into 191 rectangular segments (width 0.54 mm, thickness 0.54 mm, and lengths varying from 0.54 to 3.78 mm). The faulty code (19×19 mm²) is divided into 163 rectangular segments (width 0.85 mm, thickness 0.72 mm, and lengths varying from 0.85 to 2.57 mm). Figure 3 shows the two perpendicular codes are embedded in the solid spherical (diameter 30 mm) CAD model using the Extruded Surface function in SolidWorks. The codes are displayed in different colors in Figure 3 only for the illustration purposes. In a CAD file, these codes would be displayed in the same color, making it difficult to distinguish segments of one code from the other.

Figure 3.

Two QR codes are incorporated inside a spherical CAD model (SolidWorks^® 2016) in isometric view (the authentic code in red encodes the website link “http://engineering.nyu.edu/composites/”, while the faulty code in green encodes the word “counterfeit”).

The two codes are positioned in interpenetrating form to obfuscate their features. Such configuration also makes it possible to use certain segmented blocks to be common between both codes. Similar to the previous example, both these codes are embedded in a spherical model to compare their appearance after slicing. The angle at which these codes are oriented can be changed to the desired value and the same angle value will be needed to slice the model. In the present case, the QR code representing “counterfeit” is oriented along the default slicing direction, while the other code is oriented at 90° rotation with respect to Y-axis to demonstrate the effect. The slicing results summarized in Figure 4 show that out of the 4 testing orientations, only the 90º rotation with respect to Y-axis can successfully show the toolpath containing the authentic embedded QR code. The default setting or manual 90º rotation with respect to Z-axis show the toolpath of the embedded faulty QR code which results in printing the “counterfeit” code. Other orientations like the 90º rotation with respect to X-axis result in printing only partial and broken pieces of the code.

Figure 4.

The CAD model (SolidWorks^® 2016) containing two QR codes is processed under four different slicing orientations (CatalystEX, Stratasys, USA). Only the intended direction (Y-90°) shows the authentic QR code. Other directions show incomplete code or a mixture of segments of two codes.

The sliced part represents the geometry that will be printed by the 3D printer. However, the printing accuracy still depends on the printing technology, resolution and printing parameters. For example, a FFF technique having nozzle diameter of 100 μm will not be able to print features of 10 μm size. Hence, the designed code needs to match with the limitations of the printing technology such as printing resolution. Figure 5 demonstrates the workflow for printing the a hemispherical part with two embedded QR codes under two different orientations. Here, instead of the spherical geometry shown in Figure 4, a hemispherical geometry is used for 3D printing the actual object because of surface roughness issues at the bottom part of the sphere that impeded visualization. In the first orientation, the model is sliced under default conditions and the part is shown to retain the “counterfeit” code. On the other hand, slicing at 90° rotation with respect to Y-axis shows retaining the authentic QR code. The parts are printed using Dimension Elite™ 3D printer (Stratasys, USA) with ABSplus-grey filament (Stratasys, USA) and then CT scanned (Bruker SkyScan 1172, Micro Photonics Inc, USA) to capture the code inside as shown in the workflow in Figure 5.

Figure 5.

A complete process chain from embedding obfuscated codes CAD development (SolidWorks^® 2016), slicing and toolpath generation (CatalystEX, Stratasys, USA), to 3D printing (Dimension Elite™ FDM 3D printer, Stratasys, USA) the parts under different orientations, and to micro-CT scan (Bruker SkyScan 1172, Micro Photonics Inc, USA) that validates the embedded code types. (3D printed parts are 25.2 mm in diameter and 23.1 mm in height. The printed faulty QR code is 18×18 mm² in length and width. Each segment is 0.58 mm in width and 0.72 mm in thickness with lengths varying from 0.85 to 2.57 mm as mentioned at the beginning. The printed authentic QR code is 13.5×13.5 mm² in length and width. Each segment is 0.54 mm in width and thickness with lengths varying from 0.54 mm to 3.78 mm).

Figure 5 also shows the mciro-CT scans of the two printed orientations imported in CAD software, which show two different codes. Clarity in the scanned code images can be enhanced by using various image processing techniques. Also, depending on the CT scan resolution and the relative size of the QR code segments, small size codes may need image processing or enhancement before they can be read by QR code readers.^[48] Reading of codes may also be affected by any defects that may result from the manufacturing process.

3.2.2. Case study 2: printing with material jetting technology

In this case study, the same two QR codes are again embedded in the model. The segmentation scheme and layering design for both codes also remain the same. However, the thicknesses are re-designed as 0.030 mm for both codes to be slightly above the slicing/printing resolution (0.027 mm) with materials jetting 3D printing technology. It is expected that the QR code can be printed when the model is sliced and printed in the highest resolution direction (0.027 mm in the vertical Z direction), and not printed under other orientations that have lower resolution (most current 3D printers have a different resolution in Z direction than in X or Y direction). Figure 6 shows that the two interpenetrating codes are embedded in the cube (30 mm³) CAD model using the Extruded Boss/Base function in SolidWorks and is 3D printed using a multi-material resin printer J750 (Stratasys, USA) and the VeroClear and VeroWhite photopolymer materials (Stratasys, USA) to visually reveal the code inside. Since the code is printed in a clear polymer cube with colored material, it could be photographed from outside in this case, instead of using a micro-CT scanner. Out of the four testing orientations, the default slicing orientation (a) results in complete printing of the “counterfeit” code but missing a few sections for the authentic QR code. Rotating the cube by 45° about Z-axis (as in b) resolves this issue and prints both codes completely. On the other hand, by rotating 90° about Y-axis (as in c), the same CAD model is printed with complete authentic code but incomplete “counterfeit” code. If another 45° rotation about Z-axis (d) is performed after this, both codes can be printed completely again. Figure 6 shows that the default orientation did not result in printing a complete and readable authentic QR code. Only one orientation (c) is able to print the authentic code in full and scannable condition. Such possibilities allow embedding segmented codes that can be used against reverse engineering and counterfeit part production.

Figure 6.

A cube with two embedded obfuscating codes is 3D printed by J750 (Stratasys, USA) using PolyJet technology in VeroClear and VeroWhite photopolymer materials (Stratasys, USA) under various orientations: a. default; b. 45° rotation about Z-axis; c. 90° rotation about Y-axis; d. 90° rotation about Y-axis, then 45° rotation about Z-axis. (3D printed parts are cubes with lengths in 30 mm. The printed faulty QR code is 27.4×27.4 mm² in length and width. Each segment is 1.3 mm in width and 0.03 mm in thickness with lengths varying from 1.3 to 4 mm as mentioned at the beginning. The printed authentic QR code is 20.7×20.7 mm² in length and width. Each segment is 0.82 mm in width and 0.03 mm in thickness with lengths varying from 0.82 mm to 5.8 mm). Except for the first column, which shows a CAD model, all other images are photographs taken from a digital camera. The horizontal bright lines in the images in last columns images are due to light glare from the surface finish and are not internal features of the printed code.

4. Conclusion

Unauthorized production and reverse engineering are major challenges faced by additive manufacturing (AM) industry, especially for high value parts created in aerospace and medical industries. Existing methods do not take advantage of layer-by-layer manufacturing process of AM to develop anti-counterfeiting technologies. Therefore, an approach is developed in this work that relies on embedding two interpenetrating QR codes in the CAD model. The codes are segmented into a large number of parts and embedded as a cloud of segments (voxels), which is difficult to replicate for the purpose of counterfeiting or reverse engineering. Depending on the orientation and thickness of the embedded code segments, the slicing direction and properties need to be adjusted to obtain the correct code in the final part. Processing this file using default slicing orientation and parameters would lead to printing the faulty QR code, which can be used to identify a counterfeit part. Reverse engineering of a genuinely acquired part would not lead to the CAD file that looks identical to the original CAD file because of absence of the second code that is lost during processing of the part. These obfuscation methods can be innovatively used to develop additional strategies such that a large number of segments can be retained to act like defects in an unauthorized part, while a genuine part will be printed as fully dense. Since these methods are rooted in the design, they are expected to work across various printing technologies with appropriate adjustment for printer resolution, slice thickness and other parameters. The minimum size of the code depends on the printer resolution and slice thickness, which implies that printers with finer resolution can print codes in smaller sizes. The mecchanical properties of the parts may be affected adversely if the total part size is comparable to the embedded code size. However, in a large part, a small and segmented code may not have any effect on mechanical properites. Although the results shown in this work are only on polymer specimens, the security methods are to be implemented in the design phase and, in principle, can be applied to metallic and other materials as well. This work has not studied the effect of post-processing methods such as heat treatments of metallic parts on the embedded codes. As long as identifyable features are retained, schemes can be developed to recover the full code or develop positive identificaiton methods for the part.