Conversation Steganography (github.com)

A new tool called Conversation Steganography has been announced, enabling users to hide secret messages within seemingly innocuous chat text across various messaging platforms like WhatsApp and Signal. By leveraging a local AI model, the tool encrypts words and integrates them into natural-sounding text, allowing private conversations to evade detection from governments and surveillance systems that might flag encrypted messages as suspicious. For example, a user could type "meet me at the coffee shop at 3pm," and the tool would generate a cover text that sounds unrelated, which is then sent in the chat. This development is significant for the AI/ML community as it showcases advanced applications of steganography and local AI processing for privacy and security. Implementing AES-SIV encryption ensures both confidentiality and integrity, while the entire system operates without cloud reliance, enhancing security further. Users must follow strict protocols to maintain message integrity and need a shared secret phrase, adding a layer of complexity. Although currently a proof of concept, the potential implications for privacy in digital communications are profound, particularly in politically sensitive environments where surveillance is prevalent.