In this work, side-channel attacks (SCAs) are considered as a security metric for the implementation of hybrid cryptosystems utilizing the neural network-based Tree Parity Re-Keying Machines (TPM). A virtual study is presented within the MATLAB environment that explores various scenarios in which the TPM may be compromised. Performance metrics are evaluated to model possible embedded system implementations. A new algorithm is proposed and coined as Man-in-the-Middle Power Analysis (MIMPA) as a means to copy the TPM’s generated keys. It is shown how the algorithm can identify vulnerabilities in the physical device in which the cryptosystem is implemented by using its power emissions. Finally, a machine learning approach is used to identify the capabilities of neural networks to recognize properties of keys produced in the TPM as they are transferred to an encryption algorithm. The results show that physical exploits of TPM implementations in embedded systems can be identified and accounted for before a final release. The experiments and data acquisition is demonstrated with an implementation of a TPM-AES hybrid cryptosystem in an AVR microcontroller.

中文翻译：

---

利用侧信道发射的树型奇偶校验机实施的安全性评估

在这项工作中，边信道攻击（SCA）被视为使用基于神经网络的树奇偶校验密钥机（TPM）来实现混合密码系统的安全性度量。在MATLAB环境中提供了一个虚拟研究，探讨了可能损害TPM的各种情况。评估性能指标以对可能的嵌入式系统实现进行建模。提出了一种新算法，并将其称为中间人功率分析（MIMPA），作为复制TPM生成的密钥的一种方法。展示了该算法如何通过使用其功率发射来识别实施密码系统的物理设备中的漏洞。最后，机器学习方法用于识别神经网络的能力，以识别将TPM中生成的密钥的属性转移到加密算法时的能力。结果表明，在最终版本发布之前，可以识别并考虑嵌入式系统中TPM实现的物理漏洞。通过在AVR微控制器中实现TPM-AES混合密码系统来演示实验和数据采集。