TWC: Small: Flash Memory for Ubiquitous Hardware Security Functions

在智能手机上进行金融交易会引发许多安全问题。 银行如何确定请求是真实的？ 如何防止同一交易无意中重复？ 我们如何确保您的敏感信息不会被复制，即使手机丢失？ 强大的硬件安全功能，如设备指纹和真随机数生成器，对于解决这些问题至关重要。 然而，传统的硬件安全功能构建起来困难且昂贵。 该项目研究使用现成的闪存，这已经在大多数数字系统的今天，提供安全功能，如设备指纹，随机数发生器，和安全的信息存储。 这些安全功能将以即插即用的方式从今天的Flash中提取出来，而无需任何定制的修改，从而使几乎所有电子设备都能够实现基于硬件的安全性。 因此，该项目将在计算设备无处不在的时代大大增强安全性和隐私性。此外，该项目还将培养和教育新一代的跨学科工程师，他们既了解安全，又了解半导体器件。为了实现拟议的安全功能，该项目利用闪存的固有模拟行为，如隐藏的变化，噪声，老化等。例如，随机数可以从闪存中的热或量子噪声产生。 可以从每个存储器单元的编程/擦除定时变化中提取器件指纹，这甚至不能由闪存制造商预测或控制。 信息隐藏可以通过选择性地强调比特以产生概率差异来实现。 这种隐藏的信息将非常难以复制，甚至难以检测，除非知道特定的密钥。 这些模拟行为可以通过标准闪存接口观察到，而不会干扰正常的存储器功能。因此，所提出的安全功能将广泛适用于具有闪存的电子系统。

Performing financial transactions on a smartphone raises a number of security concerns. How can a bank be certain that a request is authentic? How do we prevent the same transaction to be unintentionally repeated? How can we ensure your sensitive information cannot be copied even if a phone is lost? Strong hardware security functions such as device fingerprints and true random number generators are essential in addressing these questions. However, traditional hardware security functions are difficult and expensive to build. This project investigates using off-the-shelf Flash memory, which is already in most digital systems today, to provide security functions like device fingerprints, random number generators, and secure information storage. These security functions will be extracted in a plug-and-play fashion from today's Flash without any customized modification, enabling hardware-based security in virtually all electronic devices. Therefore, this project will greatly enhance security and privacy in an era where computing devices are everywhere. Also, the project will train and educate a new generation of interdisciplinary engineers who can understand both security and semiconductor device.To enable the proposed security functions, this project taps into inherent analog behaviors of Flash memory such as hidden variations, noises, aging, etc. For example, random numbers can be generated from thermal or quantum noise in Flash memory. The device fingerprints can be extracted from program/erase timing variations of each memory cell, which cannot be predicted or controlled even by the Flash memory manufacturer. Information hiding can be achieved through selective stressing of bits to create probabilistic differences. Such hidden information will be very difficult to copy or even detect unless a specific secret key is known. These analog behaviors can be observed through the standard Flash memory interface without interfering with normal memory functions. Therefore, the proposed security functions will be broadly applicable to electronic systems with Flash memory.