Researchers have achieved a breakthrough in secure communication by developing an algorithm that hides sensitive information so effectively that it is impossible to detect that something is hidden.
A team led by researchers at the University of Oxford has achieved a breakthrough in secure communication by developing an algorithm that hides sensitive information so effectively that it is impossible to detect that something is hidden. Impossible.
Working closely with Carnegie Mellon University, researchers have developed a new method that could soon be widely used in digital human communication, including social media and private messaging. The ability to do so can empower vulnerable groups such as dissidents, investigative journalists and humanitarian workers.
This algorithm is applied in a setting called steganography. Steganography is a method of hiding sensitive information within harmless content. Steganography is different from encryption. Sensitive information is hidden, obscuring the fact that something is hidden. For example, a Shakespeare poem can be hidden inside an AI-generated image of a cat.
“This breakthrough in secure communication promises users an unprecedented level of security and efficiency that may enjoy a much higher range of plausible deniability than previous steganographic techniques. will work with all relevant communities of stakeholders, including humanitarian workers and investigative journalists, to unlock its potential in the real world.”
Dr. Christian Schroeder de Witt
Despite being researched for over 25 years, existing steganography approaches are generally insecure and individuals using these methods are at risk of detection. This is because previous steganography algorithms subtly altered the distribution of benign content.
To overcome this, the research team used recent breakthroughs in information theory, specifically minimum-entropy coupling. This combines her two distributions of data to maximize mutual information, but preserves individual distributions.
As a result, the new algorithm makes no statistical difference between distributing harmless content and distributing content that encodes sensitive information.
The algorithm was tested with several types of models that generate auto-generated content, including GPT-2, an open-source language model, and WAVE-RNN, a text-to-speech converter. Besides being completely secure, the new algorithm shows up to 40% higher encoding efficiency than previous steganography techniques in various applications and can hide more information within a given amount of data. This could make steganography an attractive method, even if you don’t need perfect security, due to the data compression and storage benefits.
The research team has applied for a patent for this algorithm, but plans to issue it under a free license to third parties for non-commercial responsible use. This includes academic and humanitarian use, and trusted third-party security audits. The researchers published this work as a preprint paper on arXiv and open-sourced an inefficient implementation of their method on Github. We plan to present the new algorithm at the International Conference on Learning Representations.
AI-generated content is increasingly used in normal human communication, facilitated by products such as ChatGPT, Snapchat AI stickers, and TikTok video filters. As a result, steganography may become more pervasive as the mere presence of AI-generated content arouses suspicion.
Co-author Dr. Christian Schroeder de Witt (pictured left) said: This is of great value, for example, to journalists and aid workers in countries where cryptocurrency practices are illegal. However, users should still take precautions as encryption technology may be vulnerable to side-channel attacks such as detecting steganography apps on a user’s phone. “
Contributor Professor Jakob Foerster said: It’s great to see Oxford, especially our young lab, at the forefront of everything. “
Original: New breakthrough enables fully secure digital communication
Than: University of Oxford | Carnegie Mellon University
