One Video to Steal Them All: 3D-Printing IP Theft through Optical Side-Channels Conference: Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security
Twisha Chattopadhyay, Fabricio Ceschin, Marco E. Garza, Dymytriy Zyunkin, Animesh Chhotaray, Aaron P. Stebner, Saman Zonouz, Raheem Beyah · 2025
Read One Video to Steal Them All: 3D-Printing IP Theft through Optical Side-Channels Conference: Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security by Twisha Chattopadhyay, Fabricio Ceschin, Marco E. Garza, Dymytriy Zyunkin, Animesh Chhotaray, Aaron P. Stebner, Saman Zonouz, Raheem Beyah (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-21897.txt.
Introduction
Twisha Chattopadhyay and seven co-authors show how a 3D printer's secrets can leak through video in this two thousand twenty-five security paper. Factories often watch printers with cameras, since print areas can be unsafe for people. The team shows that such footage gives away the design being printed. Their system tracks the printer nozzle as it moves. It turns that path back into the machine instructions, called G-code.
The system also recovers settings such as feed rate and extrusion rate. A checker then proves the stolen instructions match the original print. The attack threatens makers in fields from health care to defense. The paper warns that camera feeds of production lines need guarding like any other trade secret.
Catalog record
| CREATOR | Twisha Chattopadhyay, Fabricio Ceschin, Marco E. Garza, Dymytriy Zyunkin, Animesh Chhotaray, Aaron P. Stebner, Saman Zonouz, Raheem Beyah |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 80 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-21897
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
One Video to Steal Them All: 3D-Printing IP Theft through Optical Side-Channels Conference: Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security; October 13–17, 2025; Taipei, Taiwan Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security (CCS ’25), October 13–17, 2025, Taipei, Taiwan DOI: 10.1145/3719027.3744837 ISBN: 979-8-4007-1525-9/2025/10 CCS: Security and privacy Side-channel analysis and countermeasures Twisha Chattopadhyay, Fabricio Ceschin, Marco E. Garza, Dymytriy Zyunkin, Animesh Chhotaray, Aaron P. Stebner, Saman Zonouz, Raheem Beyah Abstract . The 3D printing industry is rapidly growing and is increasingly adopted in various sectors, including manufacturing, healthcare, and defense. However, the operational setup often involves hazardous environments, necessitating remote monitoring through cameras and other sensors, which opens the door to cyber-based attacks. In this paper, we show that an adversary with access to video recordings of the 3D printing process can reverse-engineer the underlying 3D print instructions. Our model tracks the printer nozzle’s movements during the printing process and maps the corresponding trajectory into G-code instructions. Further, it identifies the correct parameters such as feed rate and extrusion rate, which enable successful IP theft. To validate the success of IP theft, we design an equivalence checker that quantitatively compares two sets of 3D print instructions, evaluating their similarity in producing objects that are alike in shape, external appearance, and internal structure. Our equivalence checker, unlike other simple distance-based metrics such as normalized mean square error, is rotationally as well as translationally invariant. This is necessary to capture shifts in the base/start position of the reverse-engineered instructions relative to the actual 3D print instructions that can happen due to different camera positions. Our model achieves an average accuracy of 90.87% and generates 30.20% fewer instructions compared to the current state-of-the-art methods that produce instructions that either lead to faulty or incorrect (in terms of differences in shape and internal structure) 3D prints. Additionally, we use our model to reverse-engineer the 3D print instructions from a video recording and print a fully-functional counterfeit object. † † cc-license: by 1. Introduction The global additive-manufacturing (AM) aka 3D printing industry was valued at around $20 billion in 2023 ( GlobeNewswire 2024 ) . It is readily being adopted by different sectors considering its ability to produce complex geometries, reduce material waste, and shorten production cycles. For instance, in the healthcare sector, AM is used to print custom implants and prosthetics; the global 3D printing healthcare market is projected to reach $3.7 billion by 2026 ( Yu et al. 2023 ) . In the automotive sector, AM is used to manufacture low-cost lightweight components, prototypes, and custom parts, and enhance vehicle performance and fuel efficiency ( BMW Group 2020 ; Ford Media Center 2023 ; General Motors 2022 ) . Even governments (e.g. USA, France) are increasingly favoring the use of AM in the production of military equipment ( Astroa 2024 ; 3D Universe 2024 ) , given the compact and speedy nature of 3D printers and 3D printing respectively. In a high-valued 3D printing market, the 3D design model, typically described using an STL 1 1 1 STL stands for standard tesselation language and uses triangular meshes, typically represented as four triplets — three vertices and a norm vector — to describe the 3D model of an object. file, is the intellectual property (IP) of the designer. An IP owner customizes it with parameters that define the mechanical properties of the object that will be 3D printed. A design-compiler, commonly known as slicer ( UltiMaker 2024 ) in 3D printing, is typically used to convert the STL file with all the parameters, to form a sequence of geometric code (G-code) instructions. These instructions will be interpreted by the 3D printer firmware ( Marlin Firmware 2024 ) instruction-by-instruction to print the object additively . For example, in filament-based 3D printing ( Williams 2015 ) , an extruder heats and melts a plastic filament, which is then deposited additively layer by layer onto a print plate by the printer’s nozzle; as each layer cools and solidifies, the printer precisely adds subsequent layers, gradually forming the final 3D object. The movement of the nozzle, as well as the information on when and by how much to extrude, is defined by a sequence of G-code instructions (stored in a G-code file). Stealing the G-code IP, i.e., the G-code file, allows an adversary to produce counterfeit ( Kikuchi et al. 2018 ; Chen et al. 2019 ; Gültekin et al. 2019 ) 3D printed objects. Except for three G-code IP-theft attacks ( Do et al. 2016 ; Alkofahi et al. 2024 ; Sturm et al. 2017 ) that assume an adversary compromised either the network used to transfer the G-code and printer configurations, or the CAD/slicer software ( Sturm et al. 2017 ) , all prior works aimed at reverse-engineering the G-code IP using only access to some side channels— optical ( Liang et al. 2022 ) , acoustic ( Mativo et al. 2018 ; Chhetri et al. 2017 ; Chhetri and Al Faruque 2017 ; Al Faruque et al. 2016 ; Song et al. 2016 ) , power ( Gatlin et al. 2021 ) , thermal ( Faruque et al. 2016 ) . Since commercial 3D printing can produce hazardous emissions while melting plastic, metal, etc., these side channels are available via sensor/monitoring equipment that is used for in-situ remote quality checks. For example, there are several solutions ( Gao et al. 2018 ; 3DQue 2018 ; AstroPrint 2016 ; 3DPrinterOS 2024 ) that allow operators to use cameras — an optical side channel— to monitor the 3D printing process. Threat Model. In this work, we assume that the adversary has exploited vulnerabilities in IP-based cameras ( Liranzo and Hayajneh 2017 ) to get the video recording of the printing process, and its goal is to reverse-engineer a 3D printable G-code IP just from the video recording. See Figure 1 for a pictorial description of the threat model. Note that we do not assume that the adversary has planted any external device (e.g., smartphones ( Song et al. 2016 ) , hidden camera ( Liang et al. 2022 ) ) to gain access to the side channel. Rather, the camera is part of the 3D printing setup that is used by the operators to remotely monitor the 3D printing process, prevent material wastage, and perform post-mortem analysis in case of print failures. Additionally, we assume that the adversary has no knowledge of the original G-code, i.e., they don’t know the number of instructions the original G-code contains, the parameters that were set in order to create the G-code, the number of layers, or the ratio of instructions across layers. Further , the adversary has no knowledge regarding the position of the object being printed on the printing plate. Lastly, we assume that the video feed from the camera contains the full 3D printer at all times; however, this does not give the adversary information about the coordinate reference frame set by the operator for the printing plate. No prior works have shown that such an attack is feasible in this threat model for realistically complex parts. Moreover, no suitable metric exists to compare the similarities between the