Posts in General
Firefox Vulnerabilities: Souvenirs of Windows 95
"Mozilla doesn't have a security model for extensions and Firefox fully trusts the code of the extensions. There are no security boundaries between extensions and, to make things even worse, an extension can silently modify another extension."
Asking which of Firefox and Internet Explorer is most secure is like asking which of two random peasants is wealthier. They both might be doing their best and there may be significant differences but I wouldn't expect either to be financiers. While I'm running with this analogy, let me compare the widespread and often mandatory use of client scripts in websites (e.g., JavaScript) to CDOs: they both are designed by others with little interest in your security, they leverage your resources for their benefit, they are opaque, complex, nearly impossible to audit, and therefore untrustworthy. They have also both caused a lot of damage, as having scripting enabled is required for many attacks on browsers. How much smaller would botnets be without scripting? Like CDOs, scripting is a financial affair; it is needed to support advertising and measure the number of visitors and click-throughs. Scripting will stay with us because there's money involved, and if advertisers had their way, there would be no option to disable plugins and JavaScript, nor would there be extensions like NoScript. To be fair, there are beneficial uses for JavaScript, but it's a tangled mess with a disputable net value. Here's my take on media and advertising:
Every medium supported exclusively by advertising tends to have a net value of zero for viewers and users (viewsers?). This is where radio and TV are right now. If the value was significantly higher than zero, advertisers could wring more profits from it, for example by increasing the duration or number of annoying things, polluting your mind or gathering and exploiting more information. If it was significantly less than zero, then they would lose viewership and therefore revenue.
So, with time, and if advertising is allowed to power all websites through the requirement for scripting and JavaScript, surfing the web will become as pleasant, useful and watchable as TV, for example (with the difference that your TV can't be used --yet-- to attack people and other nations). I don't mind being locked out of websites that critically depend on advertising revenue -- just like I don't watch TV anymore because it has become a negative value proposition to me. However I mind being needlessly exposed to risks due to other people's decisions, when I use other websites. I'm looking forward to the "component directory lockdown" in Firefox 3.6 as a step in the right direction, and that's the bright light at the end of the tunnel: some things are improving.
Are We All Aware Yet?
So, here we are, in November already. We've finished up with National Cyber Security Awareness Month — feel safer? I was talking with someone who observed that he remembered "National Computer Security Day" (started back in the late 1990s) that then became "National Computer Security Week" for a few years. Well, the problems didn't go away when everyone started to call it "cyber," so we switched to a whole month but only of "awareness." This is also the "Cyber Leap Ahead Year." At the same level of progress, we'll soon have "The Decade of Living Cyber Securely." The Hundred Years' War comes to mind for some reason, but I don't think our economic system will last that long with losses mounting as they are. The Singularity may not be when computers become more powerful than the human mind, but will be the point at which all intellectual property, national security information, and financial data has been stolen and is no longer under the control of its rightful owners.
Overly gloomy? Perhaps. But consider that today is also the 21st anniversary of the Morris Internet Worm. Back then, it was a big deal because a few thousand computers were affected. Meanwhile, today's news has a story about the Conficker worm passing the 7 million host level, and growing. Back in 1988 there were about 100 known computer viruses. Today, most vendors have given up trying to measure malware as the numbers are in the millions. And now we are seeing instances of fraud based on fake anti-malware programs being marketed that actually infect the hosts on which they are installed! The sophistication and number of these things are increasing non-linearly as people continue to try to defend fundamentally unsecurable systems.
And as far as awareness goes, a few weeks ago I was talking with some grad students (not from Purdue). Someone mentioned the Worm incident; several of the students had never heard of it. I'm not suggesting that this should be required study, but it is indicative of something I think is happening: the overall awareness of security issues and history seems to be declining among the population studying computing. I did a quick poll, and many of the same students only vaguely recalled ever hearing about anything such as the Orange Book or Common Criteria, about covert channels, about reference monitors, or about a half dozen other things I mentioned. Apparently, anything older than about 5 years doesn't seem to register. I also asked them to name 5 operating systems (preferably ones they had used), and once they got to 4, most were stumped (Windows, Linux, MacOS and a couple said "Multics" because I had asked about it earlier; one young man smugly added "whatever it is running on my cellphone," which turned out to be a Windows variant). No wonder everyone insists on using the same OS, the same browser, and the same 3 programming languages — they have never been exposed to anything else!
About the same time, I was having a conversation with a senior cyber security engineer of a major security defense contractor (no, I won't say which one). The engineer was talking about a problem that had been posed in a recent RFP. I happened to mention that it sounded like something that might be best solved with a capability architecture. I got a blank look in return. Somewhat surprised, I said "You know, capabilities and rings — as in Multics and System/38." The reaction to that amazed me: "Those sound kinda familiar. Are those versions of SE Linux?"
Sigh. So much for awareness, even among the professionals who are supposed to be working in security. The problems are getting bigger faster than we have been addressing them, and too many of the next generation of computing professionals don't even know the basic fundamentals or history of information security. Unfortunately, the focus of government and industry seems to continue to be on trying to "fix" the existing platforms rather than solve the actual problems. How do we get "awareness" into that mix?
There are times when I look back over my professional career and compare it to trying to patch holes in a sinking ship while the passengers are cheerfully boring new holes in the bottom to drop in chum for the circling sharks. The biggest difference is that if I was on the ship, at least I might get a little more sun and fresh air.
More later.
Cassandra Firing GnuPG Blanks
What About the Other 11 Months?
October is "officially" National Cyber Security Awareness Month. Whoopee! As I write this, only about 27 more days before everyone slips back into their cyber stupor and ignores the issues for the other 11 months.
Yes, that is not the proper way to look at it. The proper way is to look at the lack of funding for long-term research, the lack of meaningful initiatives, the continuing lack of understanding that robust security requires actually committing resources, the lack of meaningful support for education, almost no efforts to support law enforcement, and all the elements of "Security Theater" (to use Bruce Schneier's very appropriate term) put forth as action, only to realize that not much is going to happen this month, either. After all, it is "Awareness Month" rather than "Action Month."
There was a big announcement at the end of last week where Secretary Napolitano of DHS announced that DHS had new authority to hire 1000 cybersecurity experts. Wow! That immediately went on my list of things to blog about, but before I could get to it, Bob Cringely wrote almost everything that I was going to write in his blog post The Cybersecurity Myth - Cringely on technology. (NB. Similar to Bob's correspondent, I have always disliked the term "cybersecurity" that was introduced about a dozen years ago, but it has been adopted by the hoi polloi akin to "hacker" and "virus.") I've testified before the Senate about the lack of significant education programs and the illusion of "excellence" promoted by DHS and NSA -- you can read those to get my bigger picture view of the issues on personnel in this realm. But, in summary, I think Mr. Cringely has it spot on.
Am I being too cynical? I don't really think so, although I am definitely seen by many as a professional curmudgeon in the field. This is the 6th annual Awareness Month and things are worse today than when this event was started. As one indicator, consider that the funding for meaningful education and research have hardly changed. NITRD (National Information Technology Research & Development) figures show that the fiscal 2009 allocation for Cyber Security and Information Assurance (their term) was about $321 million across all Federal agencies. Two-thirds of this amount is in budgets for Defense agencies, with the largest single amount to DARPA; the majority of these funds have gone to the "D" side of the equation (development) rather than fundamental research, and some portion has undoubtedly gone to support offensive technologies rather than building safer systems. This amount has perhaps doubled since 2001, although the level of crime and abuse has risen far more -- by at least two levels of magnitude. The funding being made available is a pittance and not enough to really address the problems.
Here's another indicator. A recent conversation with someone at McAfee revealed that new pieces of deployed malware are being indexed at a rate of about 10 per second -- and those are only the ones detected and being reported! Some of the newer attacks are incredibly sophisticated, defeating two-factor authentication and falsifying bank statements in real time. The criminals are even operating a vast network of fake merchant sites designed to corrupt visitors' machines and steal financial information. Some accounts place the annual losses in the US alone at over $100 billion per year from cyber crime activities -- well over 300 times everything being spent by the US government in R&D to stop it. (Hey, but what's 100 billion dollars, anyhow?) I have heard unpublished reports that some of the criminal gangs involved are spending tens of millions of dollars a year to write new and more effective attacks. Thus, by some estimates, the criminals are vastly outspending the US Government on R&D in this arena, and that doesn't count what other governments are spending to steal classified data and compromise infrastructure. They must be investing wisely, too: how many instances of arrests and takedowns can you recall hearing about recently?
Meanwhile, we are still awaiting the appointment of the National Cyber Cheerleader. For those keeping score, the President announced that the position was critical and he would appoint someone to that position right away. That was on May 29th. Given the delay, one wonders why the National Review was mandated as being completed in a rush 60 day period. As I noted in that earlier posting, an appointment is unlikely to make much of a difference as the position won't have real authority. Even with an appointment, there is disagreement about where the lead for cyber should be, DHS or the military. Neither really seems to take into account that this is at least as much a law enforcement problem as it is one of building better defenses. The lack of agreement means that the tenure of any appointment is likely to be controversial and contentious at worst, and largely ineffectual at best.
I could go on, but it is all rather bleak, especially when viewed through the lens of my 20+ years experience in the field. The facts and trends have been well documented for most of that time, too, so it isn't as if this is a new development. There are some bright points, but unless the problem gets a lot more attention (and resources) than it is getting now, the future is not going to look any better.
So, here are my take-aways for National Cyber Security Awareness:
- the government is more focused on us being "aware" than "secure"
- the criminals are probably outspending the government in R&D
- no one is really in charge of organizing the response, and there isn't agreement about who should
- there aren't enough real experts, and there is little real effort to create more
- too many people think "certification" means "expertise"
- law enforcement in cyber is not a priority
- real education is not a real priority
But hey, don't give up on October! It's also Vegetarian Awareness Month, National Liver Awareness Month, National Chiropractic Month, and Auto Battery Safety Month (among others). Undoubtedly there is something to celebrate without having to wait until Halloween. And that's my contribution for National Positive Attitude Month.
The Secunia Personal Software Inspector
When I made the Cassandra system years ago, I was also dreaming of something like this. It is limited to finding vulnerable software by version, not configuration, and giving links to fixes; so it doesn't help hardening a system to the point that some computer security benchmarks can. However, those security benchmarks can decrease the convenience of using a computer, so they require judgment. It can also be time consuming and moderately complex to figure out what you need to do to improve the benchmark results. By contrast, the SPI is so easy to install and use that it should be considered by anyone capable of installing software updates, or anyone managing a family member's computer. The advanced interface also pointed out that there were still issues with Internet Explorer and with Firefox for which no fixes were available. I may use Opera instead until these issues get fixed. It is unfortunate that it runs only on Windows, though.
The Secunia Personal Software Inspector is not endorsed by Purdue University CERIAS; the above are my personal opinions. I do not own any shares or interests in Secunia.
Edit: fixed the link, thanks Brett!
Odds & Ends
Cyber Leap Year Summit
I've heard from many, many people who read my blog post about this. So far, everyone who attended and was not involved with the planning of the Summit has basically agreed with my comments.
Here is an interesting post by Russ Thomas that explores the NCLY in depth from a different point of view.
Cybersecurity Legislation
There has been considerable press coverage and discussion on the intertubes about the provision in S. 773 (see my earlier post) that would allow the President to shut down critical infrastructure networks in the event of a national emergency. The people worried about the black helicopters are sure this, coupled with attempts to pass health care, are a sure sign of the Apocalypse -- or the approach of the end of the world in 2012, whichever comes first. Far less attention has been paid to other troubling aspects of the bill, such as the troubling requirement for professional certification of cyber security personnel.
According to some of the experts I have talked with, the President already has this general authority from other legislation. This simply makes it explicit. Furthermore, if we're in a declared national emergency wouldn't a centralized, coordinated response make sense? If not centered at the White House, then where else?
The bill is still in revision, although a draft of an amended version has been circulated to some groups for comment. I have been told that it is unlikely to move forward until after health care reform has been resuscitated or pronounced dead, and after the annual Federal budget appropriations process is finished. So, there may be additional issues betwixt now and then.
9/11 Comments
I wrote something in my personal blog about my 9/11 memories. It isn't really related to cyber security or Purdue, but some of my comments might be interesting to some people.
Other blog
In addition to my personal blog cited above, I also maintain a Tumbler blog with pointers to recent news items that relate to security, privacy and cyber law. It is available as <http://blog.spaf.us> (my part of the overall CERIAS blog (here) can be accessed as <http://cblog.spaf.us>). I generally post links there every day.
A Snapshot
I spent several days this week in DC, visiting officials and agencies related to cyber security. I get the sense that there is little expectation of more funding or attention in the coming fiscal year. The administration has been undergoing a bruising battle over health care, there is yet to be debate on policy for Afghanistan, and there are background engagements in constant play on issues related to the deficit. Cyber is not likely to be viewed as critical because things seem to have been going "okay" so far, and addressing cyber will be costly and require political capital. So, unless there is some splashy disaster, we might not see much progress.
ReAssure 1.20 Release
Still no sign of land
I am a big fan of the Monty Python troupe. Their silly take on several topics helped point out the absurd and pompous, and still do, but sometimes were simply lunatic in their own right.
One of their sketches, about a group of sailors stuck in a lifeboat came to mind as I was thinking about this post. The sketch starts (several times) with the line "Still no sign of land." The sketch then proceeds to a discussion of how they are so desperate that they may have to resort to cannibalism.
So why did that come to mind?
We still do not have a national Cyber Cheerleader in the Executive Office of the President. On May 29th, the President announced that he would appoint one – that cyber security was a national priority.
Three months later – nada.
Admittedly, there are other things going on: health care reform, a worsening insurgency problem in Afghanistan, hesitancy in the economic recovery, and yet more things going on that require attention from the White House. Still, cyber continues to be a problem area with huge issues. See some of the recent news to see that there is no shortage of problems – identity theft, cyber war questions, critical infrastructure vulnerability, supply chain issues, and more.
Rumor has it that several people have been approached for the Cheerleader position, but all have turned it down. This isn't overly surprising – the position has been set up as basically one where blame can be placed when something goes wrong rather than as a position to support real change. There is no budget authority, seniority, or leverage over Federal agencies where the problems occur, so there is no surprise that it is not wanted. Anyone qualified for a high-level position in this area should recognize what I described 20 years ago in "Spaf's First Law":
If you have responsibility for security but have no authority to set rules or punish violators, your own role in the organization is to take the blame when something big goes wrong.
I wonder how many false starts it will take before it is noticed that there is something wrong with the position if good people don't want it? And will that be enough to result in a change in the way the position is structured?
Meanwhile, we are losing good people from what senior leadership exists. Melissa Hathaway has resigned from the temporary position at the NSC from which she led the 60-day study, and Mischel Kwon has stepped down from leadership of US-CERT. Both were huge assets to the government and the public, and we have all lost as a result of their departure.
The crew of the lifeboat is dwindling. Gee, what next? Well, funny you should mention that.
Last week, I attended the "Cyber Leap Year Summit," which I have variously described to people who have asked as "An interesting chance to network" to "Two clowns short of a circus." (NB. I was there, so it was not three clowns short.)
The implied premise of the Summit, that bringing together a group of disparate academics and practitioners can somehow lead to a breakthrough is not a bad idea in itself. However, when you bring together far too many of them under a facilitation protocol that most of them have not heard of coupled with a forced schedule, it shouldn't be a surprise if the result in much other than some frustration. At least, that is what I heard from most of the participants I spoke with. It remains to be seen if the reporters from the various sections are able to glean something useful from the ideas that were so briefly discussed. (Trying to winnow "the best" idea from 40 suggestions given only 75 minutes and 40 type A personalities is not a fun time.)
There was also the question of "best" being brought together. In my session, there were people present who had no idea about basic security topics or history. Some of us made mention of well-known results or systems, and they went completely over the heads of the people present. Sometimes, they would point this out, and we lost time explaining. As the session progressed, the parties involved seemed to simply assume that if they hadn't heard about it, it couldn't be important, so they ignored the comments.
Here are three absurdities that seem particularly prominent to me about the whole event:
- Using "game change" as the fundamental theme is counter-productive to the issue. Referring to cyber security and privacy protection as a "game" trivializes it, and if nothing substantial occurs, it suggests that we simply haven't won the "game" yet. But in truth, these problems are something fundamental to the functioning of society, the economy, national defense, and even the rule of law. We cannot afford to "not win" this. We should not trivialize it by calling it a "game."
- Putting an arbitrary 60-90 day timeline on the proposed solutions exacerbates the problems. There was no interest in discussing the spectrum of solutions, but only talking about things that could be done right away. Unfortunately, this tends to result in people talking about more patches rather than looking at fundamental issues. It also means that potential solutions that require time (such as phasing in some product liability for bad software) are outside the scope of both discussion and consideration, and this continues to perpetuate the idea that quick fixes are somehow the solution.
- Suggesting that all that is needed is for the government to sponsor some group-think, feel-good meeting to come up with solutions is inane. Some of us have been looking at the problem set for decades, and we know some of what is needed. It will take sustained effort and some sacrifice to make a difference. Other parts of the problem are going to require sustained investigation and data gathering. There is no political will for either. Some of the approaches were even brought up in our sessions; in the one I was in, which had many economists and people from industry, the ideas were basically voted down (or derided, contrary to the protocol of the meeting) and dropped. This is part of the issue: the parties most responsible for the problem do not want to bear any responsibility for the fixes.
I raised the first two issues as the first comments in the public Q&A session on Day 1. Aneesh Chopra, the Federal Chief Technology Officer (CTO), and Susan Alexander, the Chief Technology Officer for Information and Identity Assurance at DoD, were on the panel to which I addressed the questions. I was basically told not to ask those kinds of questions, and to sit down. although the response was phrased somewhat less forcefully than that. Afterwards, no less than 22 people told me that they wanted to ask the same questions (I started counting after #5). Clearly, I was not alone in questioning the formulation of the meeting.
Do I seem discouraged? A bit. I had hoped that we would see a little more careful thought involved. There were many government observers present, and in private, one-on-one discussions with them, it was clear they were equally discouraged with what they were hearing, although they couldn't state that publicly.
However, this is yet another in long line of meetings and reports with which I have had involvement, where the good results are ignored, and the "captains of industry and government" have focused on the wrong things. But by holding continuing workshops like this one, at least it appears that the government is doing something. If nothing comes of it, they can blame the participants in some way for not coming up with good enough ideas rather than take responsibility for not asking the right questions or being willing to accept answers that are difficult to execute.
Too cynical? Perhaps. But I will continue to participate because this is NOT a "game," and the consequences of continuing to fail are not something we want to face — even with "...white wine sauce with shallots, mushrooms and garlic."
A Quick Note about Cloud Computing
I was talking to several people at the Cyber Leap Year Summit about how we have decades of research in computing that too many current researchers fail to look at because it was never put on line. We have all noticed the disturbing trend that too many papers submitted for publication do not reference anything before about 2000 -- perhaps because so much of that early work has not been indexed in search engines?
I mentioned that I had seen papers a few years back where the authors had implied that they had invented virtualization, despite the idea going back decades; at least the Wikipedia entry seems to avoid that mistake.
Someone jokingly mentioned that at least a few things were new, such as cloud computing.
Not so fast.
Some Cloud Computing is really nothing more than SaaS on virtualized platforms. That isn't new.
However, one view of Cloud Computing is that it provides seamless processing and storage somewhere on the net, where you don't have to know where it is stored, where it makes use of multiple platforms for performance and storage, and you don't need to worry about individual machine failures because the rest of the system continues forward.
Interestingly, that was precisely the goal of the distributed OS project where I did my PhD dissertation. I wrote the first prototype distributed OS kernel for the system. The name of the project? CLOUDS. The year? 1986 was when I defended, but the name was coined in 1984. (Cf. a summary article written in 1991.)
My kernel had virtual memory, process creation/deletion, object stores, capabilities, and a built-in debugger (that one could invoke after a crash -- no blue screen, simply the console message of "Shut 'er down Scotty, she's suckin mud agin.") I demonstrated it creating objects and invoking methods (actions) on them across the network on other machines, among other things. Three later PhD dissertations relied on it, as did at least 2 MS theses.
(Oh, and I wrote most of the code in VAX assembler language and it all ran on the bare hardware. I debugged it by stepping through memory, and found some hardware bugs in the process. I was a real programmer back then: i have programmed machine code on six architectures, and in over 25 other high-level languages. But I digress...)
My dissertation is not very good; I would not accept it from one of my students now, and do not recommend anyone read it. But circumstances were such that I didn't actually have an advisor for a big chunk of my research work, and the committee wanted to get me out. I never got a publication from the dissertation work, either. In retrospect, I'm not sure that was the best course of action, but I seem to have turned out more or less okay otherwise. 
Bonus item: The first Ph.D. from the group, based on an earlier attempt at the kernel, was Jim Allchin. But don't blame the Clouds group for Windows!
Bonus item: Only about 4 people ever knew, but "Clouds" was an acronym. We liked the imagery because if you combined two clouds, you simply ended up with a cloud. Up close, you couldn't tell where the boundaries of a Cloud were. And if you took some away from a cloud, you still had a cloud. Great, huh? I'm going to reveal the acronym here: Coalescing Local Objects Under Distributed Supervision. We needed the acronym for the proposal to the funding agencies, but for obvious reasons, we never referred to it as anything other than Clouds. The acronym was coined by Bill Thibault.
Bonus item: The Clouds kernel was the third OS I had written, and the final one. The second one was also in assembly language and some custom microcode for the PR1ME 500 &750 series computers (of which Georgia Tech had five). I taught a class around machine architecture and writing an OS at Georgia Tech while a grad student. I'd love to hear from anyone who took the class.
Bonus item: Although my research work quickly moved into other areas of computing, I stayed with OS long enough to help start and chair (with George Leach) the (first WEBDMS; 1989 and) SEBDMS (1991, 1992) conferences. These later evolved into the OSDI conferences -- which I have never attended. It is unlikely that many people remember this connection.
A few people still remember me for that OS work. Others know me for the work I did in mutation testing, and yet others for the work in dynamic slicing and backtracking for debugging. That was all before I started work in security and forensics. I'm to blame for more than many people know -- and I'm not telling about the rest.
But next time someone tries to tell you about their latest "new" idea, you might check with some of us older more seasoned computing folk, first, and let us reminisce about the good old days.
Beware SQL libraries missing prepared statement support
Imagine this scenario. You read that prepared statements are a good way to avoid SQL injection, because the database is given code and data explicitly and separately. You chose a database that supports prepared statements. The library you use also seems to support them as you can pass SQL code and data as two separate arguments. However... internally the library just constructs a string and sends that to the database, and doesn't use the database's prepared statement support!
An example is the library "pyPGSQL", which supports PostGreSQL in Python. It has an "execute" command taking a query and parameters as separate arguments. However, internally it constructs a string to send off (after escaping the parameters, so it *shouldn't* be vulnerable):
self.res = self.conn.conn.query(_qstr % parms)
The point is that escaping on the client side, while most likely OK, isn't as robust as letting the database handle the data separately, by using prepared statements. This particularity of pyPGSQL has been known since 2003 (forum answer). However, it's good to point it out again, as I had started writing a program using pyPGSQL, thinking that my code would be fine. It's possible that others have as well. pyPGSQL doesn't claim to support prepared statements (the absence of a "prepare" instruction should have been a clue!). Nevertheless, it surprises me that there still exist libraries not supporting prepared statements and that don't state this with an unmistakable, large warning. I found surprisingly few Python libraries supporting prepared statements:
- py-postgresql (unfortunately requires Python 3; was pg_proboscis for Python 2)
- Cristian Gafton's python-pgsql (not thread-safe)
P.S.: Note that the work-around proposed in the forum link above does not provide the security of prepared statements properly supported by a library.
P.P.S.: To clarify, I haven't demonstrated an SQL injection vulnerability in pyPGSQL. It's not about the performance penalty either. It's about escaping done by the client library (the basic implementation of bind parameters without using the database's support) being a second-rate security solution to explicitly telling the database "here is the code. Now here's the data" (prepared statements). It's about decreasing code complexity and reducing chances for "misunderstandings" (and configuration, e.g., encoding, discrepancies). It's about assurance and choosing safer technologies and architectures.
P(3)S.: Why did I expect prepared statement support? Because the Python DBI 2.0 specification for the "execute" method suggests that implementations should be using prepared statements, at least internally:
"A reference to the operation will be retained by the
cursor. If the same operation object is passed in again,
then the cursor can optimize its behavior. This is most
effective for algorithms where the same operation is used,
but different parameters are bound to it (many times)."
One more thought is that I should be more positive and congratulate the people working on Python 3 for fixing this long-standing problem. They deserve kudos!
