The 25^th ACM Symposium on Document Engineering

Despite there being many software systems that appear to detect plagiarism and AI-generated text, they do not actually work as many people suppose. Plagiarism comes in many varieties and not all are easy to detect. There are also multiple algorithms used that do not produce the same results. There are, however, interesting forensic indicators that can point to plagiarism. The software systems can be seen as a potential tool, but not as a decision system for determining plagiarism. It is, however, very easy to document some forms of plagiarism.

It is a different story for AI-generated texts. There is no proof to be found that a text was produced by a large language model, only probabilities. Depending on the use case, the amount of false positives and false negatives can also preclude the use of such systems for the detection of potential AI use. Here, too, there are forensic indicators that can show the probable use of large language models, but still cannot provide absolute proof of use.

Prof. Dr. Debora Weber-Wulff is a retired professor for Media and Computing at the University of Applied Sciences HTW Berlin in Germany. She studied applied physics at the University of California, San Diego and computer science at the University of Kiel in Germany. She received her doctorate in theoretical computer science on mechanical theorem proving at the University of Kiel. She is an active member of the working group of the Gesellschaft für Informatik (German computing society) on Ethics and Computing and Fellow of the GI. She has been researching in plagiarism since 2002 and is now using her time in retirement to discuss the use of AI in education. She published a well-received paper testing so-called AI detectors in December 2023 and has a paper on the use of AI in research currently under consideration.

When hitherto separate areas of science intersect, research opportunities tend to pop up. So it is with the fields of Document Engineering and Cybersecurity. We present an overview of certain problems that are related to these fields. This overview includes a brief summary of recent and ongoing work in our lab, which in turn includes some that has appeared at previous DocEng conferences. We summarize recent work on detecting and dealing with malicious PDF files, construction of useful malware data sets, and certain applications of tensor decomposition in the analysis of such data. We also describe some ongoing work in malware clustering, and symbolic computation as applied to offensive and defensive cyber. In recent months the topic of AI-generated documents, especially software, has created much discussion, and we will comment on this. We will conclude by pointing out certain themes in our work, as well as certain outstanding problems.

Charles Nicholas has been a faculty member at UMBC since 1988. He received the B.S. degree from the University of Michigan-Flint, and the M.S. and Ph.D. degrees from The Ohio State University, all in computer science. He has served five times as General Chair of the Conference on Information and Knowledge Management, and twice as Chair of the ACM Symposium on Principles of Document Processing. He served as chair of the department of Computer Science and Electrical Engineering at UMBC from 2003 and 2010. He is author or co-author on nearly 200 scholarly papers, and he has mentored nearly 200 M.S. students, and 20 Ph.D. students. His work has been supported by several agencies within the U.S. Department of Defense as well as a few generous corporate sponsors.