By Amit Sahai
From revolutionaries and spies to ordinary citizens living under repressive regimes, for centuries people have had the need to hide operational secrets. By an operational secret, we mean a secret that must be kept safe, where the secret is also used to perform ordinary tasks. For example, someone may want to keep the amount of cash she has on-hand secret, but she also needs to refer to this information when making everyday purchases. Even secrets like the location of hidden refugees under one’s protection could be an operational secret, since the holder of the secret would want to plan her ordinary movements to avoid the secret location that she is protecting. Through many clever means, people throughout history managed to protect such secrets. Essential to such protection was the refuge of the human mind, the ultimate sanctuary where secrets can be pondered and processed without fear of loss.
But what if this most basic assumption proves false? What if an adversary can read someone’s mind while she is thinking about the secret she needs to protect? Indeed, it is hard to imagine how someone can protect secrets when the inner dialogue of her mind can betray her. Fortunately, this scenario remains science fiction when applied to humanity. However, if we replace humans by computer programs, this situation is all too common. Computer programs, with inbuilt databases of sensitive information, are routinely captured by adversarial entities. These adversaries can reverse-engineer these programs and see every detail of how the programs “think” about their secrets as they perform their calculations. Furthermore, adversaries can even modify the programs to alter their behavior, in order to coax secrets out of the original computer code. Because computer programs with inbuilt operational secrets are so useful, cryptographic researchers have pondered this challenge as far back as the classic work of Diffie and Hellman in 1976. For decades, however, our ability to suitably “obfuscate” programs in order to defend against these kinds of attacks was based on unsatisfying approaches that failed to provide security against even moderately skilled adversaries.
This changed in 2013, due to the work of Garg, Gentry, Halevi, Raykova, Sahai and Waters, which gave the first sound mathematical approach to this problem. This work enabling a mathematical approach to securing software has been hailed as, “a watershed moment for cryptography,” by the Simons Foundation’s Quanta magazine, and its ramifications were a major theme of the Simons Institute program on Cryptography in Summer 2015. To illustrate why this recent advance in obfuscation has caused such a stir in the cryptographic community, let us consider an analogy with the ancient problem of sending encrypted messages.
CONTINUE READING