Tool type identification for forensic digital document examination
Bibliographic record
Abstract
Digital documents have become a significant part of our everyday lives. From identity documents to various legal agreements and business communications, the ability to determine the authenticity and origin of different types of documents is incredibly important. In the physical domain, this need is addressed by forensic document examiners. Although many of the analysis methods used in the physical domain do not apply in the digital realm, the forensic analysis processes in both realms still address similar objectives. In this paper, we focus on the objective of identifying the tool that created a digital document to support answering questions about the origin of a document. In contrast to many existing works on the forensic analysis of digital documents which focus on file type identification, this paper focuses on identifying the tool that is used to create a document. This is particularly relevant for forensic digital document examination (FDDE). The paper explores the use of different machine learning algorithms to analyze PDF documents to determine the tool that created the document. Given that traditional methods for digital document analysis often rely on metadata and visible content that can be tampered with, we used a structural analysis approach that builds on methods that have previously been used for file type identification. We explored the use of byte histograms and entropy measurements in developing models capable of identifying the specific software used to create PDF documents using several machine learning models. Our results showed that Convolutional Neural Networks (CNNs) outperformed other models. In further experiments, we explored the use of the same approach to identify the version of a specific tool used to create a document and alternative ways of creating PDFs from a tool. Our results confirm the feasibility of this approach for digital document tool type identification with a high level of accuracy.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".