Summary of Invited talk by Jian Guo

00:00:00 - 00:45:00

Jian Guo discusses the history of hashing, its various applications, and how it can be used to improve the efficiency and security of designs. He then presents the AES hashing competition, which was held in 2019. Guo explains how machine learning can be used to attack differential-based key recovery attacks, and how this technology has limitations. He also presents findings from previous experiences with differential-based key recovery attacks.

• 00:00:00 The speaker will discuss the evolution of cryptography security, including updates on Shatri and AES hashing. He will also provide a summary of the talk.
• 00:05:00 Jian Guo, a researcher at the Chinese Academy of Sciences, discusses the history of hashing and its various applications. He explains how hashing can be used to improve the efficiency and security of designs. Jian Guo then presents the AES hashing competition, which was held in 2019.
• 00:10:00 This 1-paragraph summary provides an overview of differential group analysis, its origins, and some of the approaches used to achieve optimum results. Differential group analysis is a major group analysis method developed in the past two decades, and includes differential pathfinding, optimization, and machine learning. Differential group analysis is used to find a good differential path between inputs and outputs, and is helpful in optimizing trade-offs between security and efficiency.
• 00:15:00 Jian Guo discussed the impact of the development of group analysis methods on the security of cryptographic systems. He explained that while the traditional methods of group analysis are still useful, the new server-based methods are more efficient and provide stronger security.
• 00:20:00 The speaker discusses the differential attack on an internal function, which begins with the initial value and fits the permutation f into the permutation. This permutation is repeated until all the message bits is processed. The squeezing phase outputs a bit at the issue iteration and here you get sufficient many bits needed. The four major instances share the same permutation as meaning that is the only difference between them. The attack framework is composed of two linear operations and the chi is the nonlinear part.
• 00:25:00 The author discussed how they have developed an AES encryption algorithm that can be attacked in six rounds. The first part of the attack is to find a good differential pass with high probability, the second part is to find the connector, and the overall goal is to have an optimum complexity. Each part of the attack can be replaced by a set of solvers, and the final step is to combine all of these steps together to achieve global optimum.
• 00:30:00 The author of the video gives a quick overview of the three key ways an attacker can break into an AES encrypted data stream. The first way is through the sub-byte substitution attack, where an attacker replaces each byte of the data with the substitution of a S-box value. The second way is through the column operation, which operates on the columns of the data. The third way is through the addition key, which is the one key used to encrypt the data. The author explains that to have a collision attack, the attacker must satisfy conditions that involve the input and output differences being equal. The author also points out that the original method for attacking AES, using the foreign key, still works even if the data being encrypted contains collisions.
• 00:35:00 This video discusses Jian Guo's research into improving the security of AES encryption by attacking it from multiple angles using dedicated software, assisted by set servers, and now machine learning. In 2019, Guo's team was able to achieve a complexity above that of human-made attacks, which means that Simon may be vulnerable to attack.
• 00:40:00 The speaker discusses how machine learning can be used to attack differential-based key recovery attacks, and how this technology has limitations. He also presents findings from previous experiences with differential-based key recovery attacks. The speaker warns of the dangers of using AES in hashing modes, as the block size offers only a limited margin of protection.
• 00:45:00 Jian Guo discusses advances in group analysis, including advances in breaking content with quantum machine learning. He notes that practical constraints such as memory limits are also a limiting factor, but notes that time is another limiting factor and that a supercomputer could theoretically be used to break all content.