Computer security depends ultimately on the security of the computer -- it's an indisputable tautology so self-evident that it seems almost insulting to point it out. Yet what may be obvious in the abstract is sometimes dangerously under-appreciated in practice. Security people come predominantly from software-centric backgrounds and we're often predisposed to relentlessly scrutinize the things we understand best while quietly assuming away everything else. But attackers, sadly, are under no obligation to play to our analytical preferences. Several recent research results make an eloquent and persuasive case that a much broader view of security is needed. A bit of simple hardware trickery, we're now reminded, can subvert a system right out from under even the most carefully vetted and protected software.

Earlier this year, Princeton graduate student Alex Halderman and seven of his colleagues discovered practical techniques for extracting the contents of DRAM memory, including cryptographic keys, after a computer has been turned off [link]. This means, among other worries, that if someone -- be it a casual thief or a foreign intelligence agent -- snatches your laptop, the fact that it had been "safely" powered down may be insufficient to protect your passwords and disk encryption keys. And the techniques are simple and non-destructive, involving little more than access to the memory chips and some canned-air coolant.

If that's not enough, this month at the USENIX workshop on Large-Scale Exploits and Emergent Threats, Samuel T. King and five colleagues at the University of Illinois Urbana-Champaign described a remarkably efficient approach to adding hardware backdoors to general-purpose computers [pdf]. Slipping a small amount of malicious circuitry (just a few thousand extra gates) into a CPU's design is sufficient to enable a wide range of remote attacks that can't be detected or prevented by conventional software security approaches. To carry out such attacks requires the ability to subvert computers wholesale as they are built or assembled. "Supply chain" threats thus operate on a different scale than we're used to thinking about -- more in the realm of governments and organized crime than lone malfeasants -- but the potential effects can be absolutely devastating under the right conditions.

And don't forget about your keyboard. At USENIX Security 2006, Gaurav Shah, Andres Molina and I introduced JitterBug attacks [pdf], in which a subverted keyboard can leak captured passwords and other secrets over the network through a completely uncompromised host computer, via simple covert timing channel. The usual countermeasures against keyboard sniffers -- virus scanners, network encryption, and so on -- don't help here. JitterBug attacks are potentially viable not only at the retail level (via a surreptitious "bump" in the keyboard cable), but also wholesale (by subverting the supply chain or through viruses that modify keyboard firmware).

Should you actually be worried? These attacks probably don't threaten all, or even most, users, but the risk that they might be carried out against any particular target isn't really the main concern here. The larger issue is that they succeed by undermining our assumptions about what we can trust. They work because our attention is focused elsewhere.

An encouraging (if rare) sign of progress in Internet security over the last decade has been the collective realization that even when we can't figure out how to exploit a vulnerability, someone else eventually might. Thankfully, it's now become (mostly) accepted practice to fix newly discovered software weaknesses even if we haven't yet worked out all the details of how to use them to conduct attacks. We know from hard experience that when "theoretical" vulnerabilities are left open long enough, the bad guys eventually figure out how to make them all too practical.

But there's a different attitude, for whatever reason, toward the underlying hardware. Even the most paranoid among us often take the integrity of computer hardware as an article of faith, something not to be seriously questioned. A common response to hardware security research is bemused puzzlement at how us overpaid academics waste our time on such impractical nonsense when there are so many real problems we could be solving instead. (Search the comments on Slashdot and other message boards about the papers cited above for examples of this reaction.)

Part of the disconnect may be because hardware threats are so unpleasant to think about from the software perspective; there doesn't seem to be much we can do about them except to hope that they won't affect us. Until, of course, they eventually do.

Reluctance to acknowledge new kinds of threats is nothing new, and it isn't limited to causal users. Even the military is susceptible. For example, Tempest attacks against cryptographic processors, in which sensitive data radiates out via spurious RF and other emanations, were first discovered by Bell Labs engineers in 1943. They noticed signals that correlated with the plain text of encrypted teletype messages coming from a certain piece of crypto hardware. According to a recently declassified NSA paper on Tempest [pdf]:

Bell Telephone faced a dilemma. They had sold the equipment to the military with the assurance that it was secure, but it wasn't. The only thing they could do was to tell the Signal Corps about it, which they did. There they met the charter members of a club of skeptics who could not believe that these tiny pips could really be exploited under practical field conditions. They are alleged to have said something like: "Don't you realize there's a war on? We can't bring our cryptographic operations to a screeching halt based on a dubious and esoteric laboratory phenomenon. If this is really dangerous, prove it."

Some things never change. Fortunately, the Bell Labs engineers quickly set up a persuasive demonstration, and so the military was ultimately convinced to develop countermeasures. That time, they were able to work out the details before the enemy did.

Thanks to Steve Bellovin for pointing me to www.nsa.gov/public/crypt_spectrum.cfm, a small trove of declassified NSA technical papers from the 1970's and 1980's, all of which are well worth reading. And no discussion of underlying security would be complete without a reference to Ken Thompson's famous Turing award lecture, Reflections on Trusting Trust.

I was part of a panel on warrantless wiretapping at last week's RSA Conference in San Francisco. The keynote session, which was moderated by the NY Times' Eric Lichtblau and which included former NSA official Bill Crowell, CDT's Jim Dempsey, attorney David Rivken, and me, focused on the role of the judiciary in national security wiretaps. Ryan Singel covered the proceedings in a brief piece on Wired's Threat Level blog.

There's a tendency to view warrantless wiretaps in strictly legal or political terms and to assume that the interception technology will correctly implement whatever the policy is supposed to be. But the reality isn't so simple. I found myself the sole techie on the RSA panel, so my role was largely to to point out that this is as much an issue of engineering as it is legal oversight. And while we don't know all the details about how NSA's wiretaps are being carried out in the US, what we do know suggests some disturbing architectural choices that make the program especially vulnerable to over-collection and abuse. In particular, assuming Mark Klein's AT&T documents are accurate, the NSA infrastructure seems much farther inside the US telecom infrastructure than would be appropriate for intercepting the exclusively international traffic that the government says it wants. The taps are apparently in domestic backbone switches rather than, say, in cable heads that leave the country, where international traffic is most concentrated (and segregated). Compounding the inherent risks of this odd design is the fact that the equipment that pans for nuggets of international communication in the stream of (off-limits) domestic traffic is apparently made up entirely of hardware provided and configured by the government, rather than the carriers. It's essentially equivalent to giving the NSA the keys to the phone company central office and hoping that they figure out which wires are the right ones to tap.

Technical details matter here. A recent paper I published with some colleagues goes into more depth about some of these issues; see a previous blog entry (linked here) about our paper.

The RSA conference itself is quite the extravaganza of commerce and technology, about as far as I care to venture from the lower-budget academic and research gatherings that comprise my natural habitat. There were something like 18,000 attendees at RSA last week, many willingly paying upwards of $2000 for the privilege. The scale is just staggeringly huge, with a level of overproduction and polish more typically associated with a Hollywood movie, or maybe a Rolling Stones concert. Some of the speakers had private dressing rooms backstage. (I didn't have one to myself, but they did lay out a nice cheese spread for us in the green room.) They made us stop at the makeup trailer before they let us on stage (I'm mildly allergic, it turns out).

If there's a silver lining under all this bling it's that digital security is unquestionably considered serious business these days. Now we just have to figure out the technical part.

I'm delighted to report that USENIX, probably the most important technical society at which I publish (and on whose board I serve), has taken a long-overdue lead toward openly disseminating scientific research. Effective immediately, all USENIX proceedings and papers will be freely available on the USENIX web site as soon as they are published [link]. (Previously, most of the organization's proceedings required a member login for access for the first year after their publication.)

For years, many authors have made their papers available on their own web sites, but the practice is haphazard, non-archivial, and, remarkably, actively discouraged by the restrictive copyright policies of many journals and conferences. So USENIX's step is important both substantively and symbolically. It reinforces why scientific papers are published in the first place: not as proprietary revenue sources, but to advance the state of the art for the benefit of society as a whole.

Unfortunately, other major technical societies that sponsor conferences and journals still cling to the antiquated notion, rooted in a rapidly-disappearing print-based publishing economy, that they naturally "own" the writings that volunteer authors, editors and reviewers produce. These organizations, which insist on copyright control as a condition of publication, argue that the sale of conference proceedings and journal subscriptions provides an essential revenue stream that subsidizes their other good works. But this income, however well it might be used, has evolved into an ill-gotten windfall. We write scientific papers first and last because we want them read. When papers were actually printed on paper it might have been reasonable to expect authors to donate the copyright in exchange for production and distribution. Today, of course, such a model seems, at best, quaintly out of touch with the needs of researchers and academics who can no longer tolerate the delay or expense of seeking out printed copies of far-flung documents they expect to find on the web.

Organizations devoted to computing research should recognize this not-so-new reality better than anyone. It's time for ACM and IEEE to follow USENIX's leadership in making scientific papers freely available to all comers. Please join me in urging them to do so.

In the month since its publication, Lawrence Wright's New Yorker profile of National Intelligence Director Mike McConnell has been widely picked apart for morsels of insight into the current administration's attitudes on the finer points of interrogation, wiretaps and privacy. I was just re-reading the piece, and, questions of waterboarding and torture aside, I was struck by this unchallenged retelling of the recent history of US cryptography regulation:

In the nineties, new encryption software that could protect telephone conversations, faxes, and e-mails from unwarranted monitoring was coming on the market, but the programs could also block entirely legal efforts to eavesdrop on criminals or potential terrorists. Under McConnell's direction, the N.S.A. developed a sophisticated device, the Clipper Chip, with a superior ability to encrypt any electronic transmission; it also allowed law-enforcement officials, given the proper authority, to decipher and eavesdrop on the encrypted communications of others. Privacy advocates criticized the device, though, and the Clipper was abandoned by 1996. "They convinced the folks on the Hill that they couldn't trust the government to do what it said it was going to do," Richard Wilhelm, who was in charge of information warfare under McConnell, says.

But that's not actually what happened. In fact, almost everything about this quote -- the whos, the whys and the whens -- is not merely slanted in its interpretation, but factually contradicted by the public record.

First, the Clipper Chip itself was abandoned not because of concerns about privacy (although it certainly became a lightning rod for criticism on that front), but rather because it was found to have serious technical vulnerabilities that had escaped the NSA's internal review process and that rendered the system ineffective for its intended purpose. I discovered and published in the open literature the first of these flaws at the ACM CCS conference in 1994, about a year after Clipper's introduction. (Technically inclined readers can see Protocol Failure in the Escrowed Encryption Standard [pdf link] for details.)

More generally, while Congress (presumably the Hill on which the folks referred to above resided) held several hearings on the subject, it never actually enacted (or even voted on) any cryptography legislation during the 1990's. Instead, the relaxation of cryptography export regulations (the lever through which the government controlled US encryption software and hardware) was entirely a product of the executive branch. Department of Commerce export regulations that took effect at the beginning of 2000 effectively eliminated the major controls on mass-market and open source encryption products, and in 2004 the rules were relaxed even further. In other words, the current policy promoting widely available encryption has been actively supported not just by the previous administration, but by the present one as well.

Government attempts in the 1990's to control cryptography were ultimately doomed not by some kind of weak-kneed policy capitulation (even if some in the national security community might now find it convenient to portray it as such), but by inexorable advances in technology (particularly what Larry Lessig calls "west coast code"). Even before the hardware-only Clipper program was announced, the performance of software-based encryption on general purpose computers had become competitive with that of expensive purpose-built encryption chips. At the same time, growth in the largely open-source-based networked computing world increasingly depended on reliable and inexpensive communications security, something fundamentally incompatible with the key escrow scheme underlying Clipper. Coupled with the inherent risk that a central encryption key database would itself become a target for foreign intelligence agents, it became apparent to just about everyone by the end of the decade that decontrol was the only viable policy option.

Tellingly, even after the attacks of September 11th, 2001 the Bush administration made no serious attempt to control or otherwise reign in the public availability or use of strong encryption. Indeed, its only action on cryptography policy came in 2004, when it further liberalized the few remaining export controls on the technology.

It's unfortunate that the New Yorker failed to check the facts about Clipper, but that's not the central concern here. Intelligence agencies (and their officials) largely operate under a thick veil of legally enforced secrecy, carrying with it a special obligation to be scrupulously truthful in their analysis. Should we trust them to be more forthright than the former official quoted in the excerpt when the facts are not so easily checked? We often have no choice.

When I published my Clipper paper in 1994, the NSA -- to its great credit at the time -- immediately acknowledged the problems, despite the embarrassment the whole affair no doubt caused them. And that is precisely how we must expect our intelligence agencies to behave. Intelligence spun to serve a political narrative isn't intelligence, it's just politics.

Photo above: Mykotronx MYK-78T Clipper Chip, as installed in an AT&T TSD-3600E encrypting telephone. More Clipper photos here.

I'll be giving a talk tomorrow (Thursday 14 February) about the review of electronic voting systems we conducted last fall here at U. Penn. For those in the Philadelphia area, the talk is on the Penn Campus in the Levine Hall Auditorium from 3:00-4:15pm. Sorry for not getting this up on the blog earlier.

A recurring theme in this blog over the last year has been how the sweeping surveillance technology envisioned by the 2007 US Protect America Act introduces fundamental technical vulnerabilities into the nation's communications infrastructure. These risks should worry law enforcement and the national security community at least as much as they worry civil liberties advocates. A post last October mentioned an analysis that I was writing with Steve Bellovin, Whit Diffie, Susan Landau, Peter Neumann and Jennifer Rexford.

The final version of our paper, "Risking Communications Security: Potential Hazards of the Protect America Act," will be published in the January/February 2008 issue of IEEE Security and Privacy, which hits the stands in a few weeks. But you can download a preprint of our article today at https://www.mattblaze.org/papers/paa-ieee.pdf [PDF]. Remember, you saw it here first.

This week the Senate takes up whether to make the warrantless wiretapping act permanent, so the subject is extremely timely. The technical risks created by unsupervised wiretapping on this scale are enormously serious (and reason enough to urge Congress to let this misguided law expire). But safety and security may be subordinated in the debate to more entrenched interests. In particular, AT&T and other communications carriers have been lobbying hard for an amendment that gives them retroactive immunity from any liability they might have had for turning over Americans' communications and call records to the government, legally or not, before the act passed.

As someone who began his professional career in the Bell System (and who stayed around through several of its successors), the push for telco immunity represents an especially bitter disillusionment for me. Say what you will about the old Phone Company, but respect for customer privacy was once a deeply rooted point of pride in the corporate ethos. There was no faster way to be fired (or worse) than to snoop into call records or facilitate illegal wiretaps, well intentioned or not. And it was genuinely part of the culture; we believed in it, even those of us ordinarily disposed toward a skeptical view of the official company line. Now it all seems like just another bit of cynical, focus-group-tested PR.

Apologies to David Byrne for paraphrasing his hauntingly prescient lyrics. And thanks to Eric Cronin for discovering and scanning the now sadly nostalgic reminder reproduced at right.

I've dabbled with photography for most of my life, defiantly regressing toward bulkier equipment and more cumbersome technique as cameras become inexorably smaller and easier to use with each successive generation. Lately, I've settled on a large format camera with a tethered digital scanning back, which is about as unwieldy as I'd likely want to endure, at least without employing a pack mule.

But for all the effort I'm apparently willing to expend to take pictures, I'm pathologically negligent about actually doing anything with them once they've been taken. Scattered about my home and office are boxes of unprinted negatives and disk drives of unprocessed captures (this is, admittedly, overwhelmingly for the best). So it was somewhat against my natural inclinations that a few years ago I started making available an idiosyncratic (that is to say, random) selection of images elsewhere on this web site.

Publishing photos on my own server has been a mixed bag. It maximizes my control over the presentation, which appeals to the control freak in me, but it also creates lot of non-photographic work every time I want to add a new picture -- editing html, creating thumbnails, updating links and indexes. I've got generally ample bandwidth, but every now and then someone hotlinks into an image from some wildly popular place and everything slows to a crawl. I could build better tools and install better filters to fix most of these shortcomings, but, true to my own form I suppose, I haven't. So a few weeks ago I finally gave in to the temptations of convenience and opened an account on Flickr.

I'm slowly (but thus far, at least, steadily) populating my Flickr page with everything from casual snapshots to dubiously artistic abstractions. It's still an experiment for me, but a side effect has been to bring into focus (sorry) some of what's good and what's bad about how we publish on the web.

Flickr is a mixed bag, too, but it's different mix than doing everything on my own has been. It's certainly easier. And there's an appeal to being able easily link photos into communities of other photographers, leave comments, and attract a more diverse audience. So far at it seems not to have scaled up unmanageably or become hopelessly mired in robotic spam. They make it easy to publish under a Creative Commons license, which I use for most of my pictures. But there are also small annoyances. The thumbnail resizing algorithm sometimes adds too much contrast for my taste -- I'd welcome an option to re-scale certain images myself. The page layout is inflexible. There's no way to control who can view (and mine data from) your social network. And they get intellectual property wrong in small but important ways. For example, every published photo is prominently labeled "this photo is public", risking the impression that copyrighted and CC-licensed photos are actually in the public domain and free for any use without permission or attribution.

I wonder how most people feel about that; do we think of our own creative output, however humble, as being deserving of protection? Or do we tend to regard intellectual property as legitimately applying only to "serious" for-profit enterprises (with Hollywood and the music industry rightfully occupying rarefied positions at the top of the socio-legal hierarchy)? A couple of years ago I led a small discussion group for incoming students at my overpriced Ivy-league university. The topic was Larry Lessig's book Free Culture, and I started by asking what I thought was a softball question about who was exclusively a content consumer and who occasionally produced creative things deserving protection and recognition. To my surprise, everyone put themselves squarely in the former category. These were talented kids, mostly from privileged background -- hardly a group that tends to be shy about their own creativity. Maybe it was the setting or maybe it was how I asked. But perhaps their apparent humility was symptomatic of something deeply rooted in our current copyright mess. Until we can abandon this increasingly artificial distinction -- "consumers" and "content producers", us against them -- our lopsided copyright laws seem bound to continue on their current draconian trajectory, with public disrespect for the rules becoming correspondingly all the more flagrant.

Today Ohio Secretary of State Jennifer Brunner released the results of a comprehensive security review of the electronic voting systems used in her state. The study was similar in scope to the California top-to-bottom review conducted this summer (with which I was also involved), covering the systems used in Ohio. The project contracted several academic teams and others to examine the election procedures, equipment and source code used in that state, with the aim of identifying any problems that might render elections vulnerable to tampering under operational conditions.

The ten-week project examined in detail the touch-screen, optical scan, and election management technology from e-voting vendors ES&S, Hart InterCivic and Premier Election Systems (formerly Diebold). Project PI Patrick McDaniel (of Penn State) coordinated the academic teams and led the study of the Hart and Premier Systems (parts of which had already been reviewed in the California study). Giovanni Vigna (of WebWise Security and UCSB) led the team that did penetration testing of the ES&S system.

I led the University of Pennsylvania-based team, which examined the ES&S source code. This was particularly interesting, because, unlike Hart and Premier, the full ES&S source code suite hadn't previously been studied by the academic security community, although ES&S products are used by voters in 43 US states and elsewhere around the world. The study represented a rather unique opportunity to contribute to our understanding of e-voting security in practice, both inside and outside Ohio.

My group -- Adam Aviv, Pavol Cerny, Sandy Clark, Eric Cronin, Gaurav Shah, and Micah Sherr -- worked full-time with the source code and sample voting machines in a secure room on the Penn campus, trying to find ways to defeat security mechanisms under various kinds of real-world conditions. (Our confidentiality agreement prevented us from saying anything about the project until today, which is why we may have seemed especially unsociable for the last few months.)

As our report describes, we largely succeeded at finding exploitable vulnerabilities that could affect the integrity of elections that use this equipment.

The report is long and detailed, and speaks for itself far better than I can here. A brief statement from Patrick McDaniel and me can be found (PDF format) here. Our full 334 page report can be downloaded (11MB, PDF format) from the Ohio Secretary of State's web site at http://www.sos.state.oh.us/SOS/upload/everest/14-AcademicFinalEVERESTReport.pdf .

There were other parts to the study (called "Project EVEREST") than just the source code analysis, and, of course, there is also the question of how to actually secure elections in practice given the problems we found. The Ohio Secretary of State's web site [link] has a nice summary of the review and of the Secretary's recommendations.

Addendum (4 June 2008): Apparently the Ohio SoS web site was recently reorganized, breaking the original links in this post. I've updated the links to point to their new locations.

Addendum (24 March 2009): See this post for a recent case of vote stealing that exploited a user-interface flaw in the ES&S iVotronic.

Most of the debate surrounding the Protect America Act -- hastily passed this summer and eliminating requirements for court orders for many kinds of government wiretaps -- has focused on the perceived need to balance national security against individual privacy. And while it's surely true that wiretap policy strikes at the heart of that balance, the specter of unrestrained government spying may not be the the most immediate reason to fear this ill-conceived and dangerous change in US law. Civil liberties concerns aside, the engineering implications of these new wiretapping rules, coupled with what we can discern about how they are being implemented, should be at least as unsettling to hawks as to doves.

My colleagues Steve Bellovin, Whit Diffie, Susan Landau, Peter Neumann, Jennifer Rexford and I just completed a brief analysis of the Protect America Act from the networking, surveillance and security engineering perspective. Our draft technical report, Risking Communications Security: Potential Hazards of the "Protect America Act" can be found at www.mattblaze.org/papers/paa-comsec-draft.pdf [pdf format].

Update 10/13/07: A less-technical version can be found at www.mattblaze.org/papers/paa-briefing.pdf [pdf format].

Update 10/22/07: A new (and slighty less drafty) draft of the full technical report is now online, at www.mattblaze.org/papers/paa-comsec-draft.pdf [pdf format].

This morning the Electronic Frontier Foundation will be releasing the first installment of many hundreds of scanned pages concerning the FBI's DCS-3000 surveillance Data Collection System. Most of the documents, which were obtained through a Freedom of Information Act (FOIA) lawsuit filed by the EFF, are heavily (almost laughably) redacted user manuals and internal FBI memos, with many pages almost entirely devoid of text. Nonetheless, what remains provides a rare, if fragmented and cryptic, glimpse of the state of FBI surveillance technology in general and CALEA wiretapping in particular.

Ryan Singel has a nice summary of the documents in his article on Wired's web site. It will take time to digest the huge volume of incomplete and jargon-laden material in the documents, but interesting tidbits emerge even from a cursory reading.

Contrary to previous speculation, DCS-3000 is much more than an updated version of the FBI's Carnivore Internet interception and collection device first disclosed seven years ago. Instead, the DCS system appears to be a comprehensive software suite for managing and collecting data from a variety of Title III (law enforcement) surveillance technologies, including Internet wiretaps, wireline voice telephony, cellular, "push-to-talk", and maybe others. The system provides a single interface for managing and collecting evidence from all the different kinds of wiretaps the FBI uses, connected via a "DCSNet" for getting tapped traffic to any FBI field office in the US. There are references to several other FBI systems as well, most notably the Bureau's ill-fated Trilogy case management system, and also something called DCS-5000, which is described as an analogous system for managing FISA (national security) taps. The software is definitely large and complex -- there are mentions of multi-week training courses for the agents who use it.

That complexity itself raises some difficult security questions. As my colleague Steve Bellovin points out, the new documents suggest that the FBI may have failed to adequately secure the system against an insider threat. But aside from the usual risks that the software could be subverted or abused, in a wiretapping system there's also the problem of ensuring that intercepted evidence is faithfully recorded. And that, it turns out, can be harder than it sounds.

Two years ago, my graduate students and I discovered basic flaws in the in-band signaling mechanisms used for many years in older analog voice telephone wiretaps. The flaws allow a wiretap target to interfere with a phone tap by playing special tones that cause interception equipment to shut down prematurely or record misleading call data. We speculated, based on the documents available to us then, that the CALEA-based interception system now used by the FBI might suffer from similar problems. The FBI denied this at the time, claiming that only a few systems remain vulnerable to our attacks. But sure enough, the EFF's new documents refer in several places to continued support for in-band "C-tone" signaling in voice line taps (for example, on page 53 of this pdf document). No doubt, these features were included to provide backward compatibility with older equipment. And the result is backward compatibility with older bugs.