Ironclad Apps: End-to-End Security via Automated Full-System Verification

Re:Ironclad Apps: End-to-End Security via Automated Full-Sys... by kevin_b_er on Friday November 21, 2014 @02:17PM (#8642799)

TPM and Remote Attestation. This is an article on Microsoft working hard on DRM schemes.

> An Ironclad App enables a user to securely transmit her data to a remote machine with the guarantee that ev-ery instruction executed on that machine adheres to a formal high-level specification.

... ...

> Today, when Alice submits her personal data to a remote service, she has little assurance that her data will remain secure. At best, she has vague legal guarantees provided by the service’s privacy policy and the hope that the owner will follow industry best practices to secure her data. Even then, a vulnerable OS, library, or application may under-mine the service provider’s best intentions [46].

If we presume the user is a DRM-demanding media company, the translation is easy to a member of the public:

> An Ironclad App enables a media company to securely transmit data to a user's machine with the guarantee that every instruction executed on the user's machine adheres to a formal high-level specification.

... ...

Today, when Alice submits her copyrighted data to a user's machine, she has little assurance that her copyrighted data will remain secure. At best, she has vague legal guarantees provided by the service’s privacy policy and the hope that the user will follow industry best practices to secure her copyrighted data. Even then, a vulnerable OS, library, or application may under-mine the digital restrictions management.

[ Parent | on HN ]

Re:Ironclad Apps: End-to-End Security via Automated Full-Sys... by jameshart on Friday November 21, 2014 @03:11PM (#8643145)

That this would apply to DRM was my first thought as well, though I think it's a little disingenuous to imply that the use of a personal-privacy-preserving scenario instead of a DRM scenario as the motivating narrative for the research might in some way be an attempt to whitewash the technology. Microsoft genuinely has an interest in solving the problem of, for example, how to create trust for Word users to store their documents in Microsoft's cloud, as well as an interest in how to enable major league sports to trust Windows phone devices to stream live games. Trust's a big deal.

The fact is that the same things which make DRM impossible absolutely do make it impossible for users to trust file sharing websites, or cloud-compute customers to trust platform-as-a-service providers. And this capability has the potential to support all kinds of scenarios - some arguably good from a personal-privacy-and-liberty point of view, some maybe bad from your subjective viewpoint.

The DRM narrative always has this 'big media hanging on to their outdated business models' narrative, but often the same people who complain about intrusive DRM are the first to complain when some company abuses an individual's copyright by using a flickr photo without permission. DRM isn't evil, just because some people might use it in ways you disapprove of.

DRM definitely has the potential to do good things. The DRM-based Steam marketplace (and also to a large extent the mobile appstores) has probably been a net good, for example, for both video game creators and consumers.

[ Parent | on HN ]

Re:Ironclad Apps: End-to-End Security via Automated Full-Sys... by jude- on Friday November 21, 2014 @03:54PM (#8643324)

FWIW, their threat model assumes that Alice can't subvert the software with hardware-based attacks. Relying on this system for DRM would be pretty shortsighted, since (1) Alice obviously has physical access to her hardware, and (2) Alice could emulate the hardware in software, and mount "physical" attacks against it without any dedicated equipment or electrical engineering know-how. A simple screen-grabber would circumvent this.

[ Parent | on HN ]

Re:Ironclad Apps: End-to-End Security via Automated Full-Sys... by kevin_b_er on Friday November 21, 2014 @05:14PM (#8643695)

That point of remote attestation. If you have the hardware module signs all software and the processor, coupled with an unblockable secure mode that only runs if the TPM authorizes it, then you have created a secure zone in the processor of which only TPM-authorized software can run. This defeats (1) and (2), with (2) having the obvious exception of human-observable components being observable anyways. This does serve DRM nicely. They don't care if you put a camera in front of a monitor for the purposes of DRM.

They coupled this with an OS and driver that is formally verified to eliminated security bugs.

[ Parent | on HN ]

Re:Ironclad Apps: End-to-End Security via Automated Full-Sys... by rictic on Friday November 21, 2014 @06:14PM (#8643957)

The DRM use case of TPM is mostly security theater, as after all that decoding work you still need to send the bits over an interface to speakers / a display, and even if that interface is secured there's no way to determine that there's not a studio quality microphone / video camera on the other side of an airgap.

[ Parent | on HN ]