My humble understanding is this is a very theoretical result, unlikely to result in unreversable malware, or improvements in the DRM near you.<p>Here's a brief summary (obviously, I might be missing a lot of things):<p>1. "They (researchers in 2001, some of which are authors of this new paper) showed that there exist unobfuscatable functions – a family of functions {f s } such that given any circuit that implements f s , an efficient procedure can extract the secret s; however, any efficient adversary given only black-box access to f s cannot guess even a single bit of s with non-negligible advantage."<p>That result still holds - one cannot obfuscate any function, and this is proven.<p>2. "indistinguishability obfuscation: An indistinguishability obfuscator iO for a class of circuits C guarantees that given two <i>equivalent</i> circuits C 1 and C 2 from the class, the two distribution of obfuscations iO(C 1 ) and iO(C 2 ) should be computationally indistinguishable."<p>Note that this works <i>only for equivalent circuits</i>.<p>3. "Using indistinguishability obfuscator for NC 1 together with any (leveled) fully homomorphic encryption (FHE) scheme with decryption in NC 1 (e.g. [Gen09b, BV11, BGV12, Bra12, GSW13]), we show how to obtain an indistinguishability obfuscator for all polynomial-size circuits".<p>Again, this is indistinguishability obfuscator, which works only for equivalent circuits. Also, FHE is <i>very</i> slow nowadays, AFAIK there are no actual deployments of that concept, because of the prohibitive slowness (e.g. a single AES encryption taking days).<p>4. "Using indistinguishability obfuscator for polynomial-size circuits, together with injective one-way functions, public-key encryption, and a novel variant of Sahai’s simulation-sound non-interactive zero knowledge [Sah99] proofs, we show how to obtain functional encryption schemes supporting all polynomial-size circuits."<p>This is awesome and sounds like it <i>can</i> obfuscate malware or be used to make actual DRM, but again, the indistinguishability obfuscator is likely so slow as to not be practical these days. Maybe in a few decades?<p>Obviously I'm not writing this to take anything away from this huge theoretical result - just saying this is likely not what other commenters think it is. And again, my reading of this is very possibly inaccurate.