Imago Obscura: An Image Privacy AI Co-pilot to Enable Identification and Mitigation of Risks

TL;DR

Imago Obscura tackles the challenge of balance between privacy and publicity in image sharing by introducing a human–AI copilot that helps users identify contextually relevant privacy risks and apply obfuscation techniques. The approach is grounded in a formative study with image-editing experts to derive five design requirements, implemented as a Krita plugin that leverages a multimodal AI ensemble and theory-informed prompting. Evaluation with 15 end-users shows increased awareness, motivation, and ability to mitigate risks while preserving sharing intent, supporting a more informed privacy trade-off. The work highlights the value of scaffolded prompting and human-in-the-loop design for usable privacy protection in image sharing, while noting guardrails and ethical considerations for generative obfuscation. Overall, Imago Obscura demonstrates a practical pathway for deploying AI-assisted privacy tools that empower users without sacrificing authenticity or control.

Abstract

Users often struggle to navigate the privacy / publicity boundary in sharing images online: they may lack awareness of image privacy risks and/or the ability to apply effective mitigation strategies. To address this challenge, we introduce and evaluate Imago Obscura, an AI-powered, image-editing copilot that enables users to identify and mitigate privacy risks with images they intend to share. Driven by design requirements from a formative user study with 7 image-editing experts, Imago Obscura enables users to articulate their image-sharing intent and privacy concerns. The system uses these inputs to surface contextually pertinent privacy risks, and then recommends and facilitates application of a suite of obfuscation techniques found to be effective in prior literature -- e.g., inpainting, blurring, and generative content replacement. We evaluated Imago Obscura with 15 end-users in a lab study and found that it greatly improved users' awareness of image privacy risks and their ability to address those risks, allowing them to make more informed sharing decisions.

Figures (13)

• Figure 1: Overview of our methodology. We conducted a formative study to derive design requirements, built a tool based on those requirements, and evaluated it with end-users using their personal photos.
• Figure 2: Imago Obscura enables users to express their sharing intent and their privacy concerns in natural language, subsequently identifying pertinent risks and recommending obfuscation techniques.
• Figure 3: Imago Obscura enables the user to directly select areas of concerns which the tool will automatically precisely select and highlight in green, subsequently identifying pertinent risks and recommending obfuscation techniques.
• Figure 4: Imago Obscura addresses "self disclosure risks". (1) Identifies that the numbered candle can reveal personal information. (2) Recommends removing the candle from the image. (3) Precisely selects the sensitive area, the candle, and applies inpainting.
• Figure 5: Imago Obscura addresses "identity exposure risk". (1) Identifies that the tattoo can reveal the person's identity. (2) Recommends to replace the tattoo with a new one. (3) Precisely selects the sensitive area, the tattoo, and applies generative content replacement.