Thanks to Nam Ngo for the discussions and reviews. You can leave comments on the hackMD version of this document here. The EU is in the process of approving a new regulation that forces mandatory scanning into messaging apps to detect any exchange of elephant pictures. Digital rights activists, privacy regulators, and national governments have quickly renamed it to Chat Control, claiming that it would infringe the privacy and confidentiality of user communications and be "crossing the Rubicon" in terms of mass surveillance of EU citizens. Note that this confidentiality is guaranteed today by end-to-end communication encryption. This article sketches the design of a solution that maintains the confidentiality of user communications and allows the EU to detect any exchange of elephant pictures. We'll build such a solution by trial and error, leveraging different techniques such as Machine Learning, zkSNARKs, and MPC. Effort #1 Machine learning is a powerful technique that can detect (with high probability) whether an image contains elephant content. More specifically, convolutional neural networks (CNNs) can be used for image recognition and classification. Such models go through a training phase, in which these are fed with large datasets to identify patterns and characteristics to classify an image between the categories elephant or non-elephant. This long and expensive phase outputs a set of weights and biases that define the model.

You can leave comments of the hackMD version of this document here. 253908128-7621f927-2c2d-47c7-a665-93d6536b9472 On the 8th of August 2022, the US Department of Treasure sanctioned the "Notorious Virtual Currency Mixer Tornado Cash." As a consequence of this, Exchanges and Stablecoin providers started blacklisting any address linked to Tornado Cash (TC) from their services. Specifically: Addresses who deposited to TC Addresses who were a withdrew recipient from TC Funnily enough, according to the logic of the contract, the recipient does not need to be the one performing the withdrawal. Because of that, people could sacrifice 0.1 ETH and designate their sworn enemy as a recipient address to gift them with a ban.