 Hello, everyone. My name is Jia Huihe. In this video, I will briefly introduce our work named Straighten Cube Attacks, Improved Methods to Recover Massive Super Parties. This is a joint work with Kai Hu, Bart Praniel, and Mei Qin Wang. The cube attack was proposed by Dina and Shamier at EuroCraft 2009. Any outpot beat of a symmetric cipher can be expressed as a boolean function of the secret variables k and the public variables x. The coefficient of a cube term x2u in this boolean function is called the superpoly. Whose value can be calculated according to a Mobius transform? In the online phase of the cube attack, the attacker can fix the non-cube variables to constants and recover the exact expression of the superpoly. Then in the online phase, he can calculate the value of the superpoly and establish an equation for k. Some information of k can be extracted by solving the equation. In the early stage of the cube attack, only linear or quadratic superpolys are targeted. Later, thanks to the introduction of the division property and the MRP modeling method, even complex superpolys can be recovered practically. The division property can also be revisited from an algebraic perspective, leading to the so-called monomial prediction technique. At AsiaCraft 2021, Hu Etou proposed a framework called nested monomial predictions by combining the divide and conquer strategy with the monomial prediction technique. Our work is a natural follow-up to their work. For composite boolean functions setting x to y, the monomial prediction or division property allows to predict if x2u appears in the nf or y2v by counting the number of so-called monomial or division trials. For a cube term x2u, if we can determine all possible gkx2u contained by f, we can determine the superpoly accordingly. From our perspective of structure, the nested monomial predictions consist of two components, which we call the coefficient solver and the term expander. The coefficient solver is responsible for computing the superpoly for each term within a time limit, while the term expander is responsible for expanding all solver terms into terms of a deep round. Since two parts are round iteratively, until no answer determines the remain. We experimentally found that the efficiency of the whole framework is mainly determined by the efficiency of the coefficient solver. The NP chooses the monomial prediction as its coefficient solver, but counting the number of monomial trials at once will become impractical as the number of rounds grows. Therefore, we redesigned a two-stage coefficient solver, where the first stage is considered the most time consuming part. The remaining problem is how to model the two stages. So the monomial prediction can also be used to characterize the first stage further analysis shows that it has a problem of over-precision. These things bias us to sacrifice some accuracy for efficiency, leading to a new technique called core monomial prediction. By combining CMP with NP, we can derive a simpler MLP model for the first stage. This new MLP model greatly speeds up the superpoly recovery. As a result, we not only verified the previous results at a much lower time cost, but also recovered new superpoly for more rounds of several string surface. Finally, by enhancing the key recovery method based on the mobius transform, we also show how to extract one-bit information of the key bits from a massive superpoly containing a proof symmetry 2 to the 30.5 terms. Thanks for your attention. If you have any question, you can contact me via the following email.