DEDALUS & ICARUS: Image Privacy Classification Systems with Risk Oriented Explanations
Bibliographic record
Abstract
The rise of online image sharing raises significant privacy concerns, as users may inadvertently disclose sensitive personal information. This paper introduces two Artificial Intelligence-driven systems, DEDALUS and ICARUS, designed to enhance user awareness and information security. DEDALUS is a modular image privacy classification system with advanced explainability capabilities. Its classification module, based on a voting ensemble of vision models, achieves good performance on an extended version of the PrivacyAlert dataset. Its explainability module combines a Large Language Model (LLM)-based risk assessment with LIME to interpret visual model decisions. To illustrate how DEDALUS’s capabilities can be integrated into a broader privacy-preserving application, we also introduce ICARUS, a system specialized in detecting privacy-related disclosures. It uses a vision model trained on a custom dataset to identify images containing Personally Identifiable Information (PII), and integrates the DEDALUS classifier via API chaining to issue user-friendly warnings. Two proof-of-concept applications are proposed: DEDALUS Workbench, for model analysis and dataset creation, and ICARUS Watcher, which simulates image sharing in unsecured environments. The ICARUS API, used in the Watcher demo, monitors image-sharing activity and issues alerts for both PII detection and private image classification. Together, these systems contribute to computer vision, AI ethics, and digital privacy by offering a novel approach to mitigating the risks associated with sharing sensitive visual content.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.005 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".