Wednesday, November 30, 2016

The event horizon of software liability and cyber insurance

Software liability and cyber insurance seem inevitable but you can never reach them - they are singularities.

There's a gravity in the policy world to try to "solve systemic information security risk" via one of two horrible ideas:

  • Cyber Insurance
  • Software Liabilities

These twin black holes spin around each other, generating gravity waves that can be felt from every other part of the information security universe.

The latest musing into this quixotic adventure is Rob Knake's idea to have the Federal Govt backstop universal cyber insurance - eventually leading to massive SEC-level controls over every company in America:
There are not good ideas. Also, email-spoofing is not what anyone does when it comes to phishing in 2016 - which is a weird technical detail to have in this paper at all.
As much as AIG would love to be the middleman in a massive new insurance market for which we have no actuarial data, but where the risk is pushed onto the US Taxpayer , the reality is there are some risks you cannot insure. Insurance was created during the Great Fire of London, but fire does not choose to burn down only the houses of the insured to cause maximum damage to the taxpayer the way a cyber adversary would. This system would be built to create an additional vulnerability on the state that another state could take advantage of.

From a technical perspective, the idea is also bankrupt. As Rob himself points out, we don't know what WORKS when it comes to securing things, and even if we knew what worked in the past, we would not know that it would continue to work in the future.
The smart thing to do is not try to build a new, trusted email, but just not to trust email. I don't know why Knake is so hot on email spoofing. Also, I want to point out that when an APT does their job right, you never know you took damage. What exactly are we insuring?

And yet, you have seen a burgeoning market for security products which offer guarantees, often backstopped by insurance companies who treat it like a marketing wager, such as this one by Cymmetria. In this end, this may be as "good as we get" when it comes to how insurance is going to work in this space.

The following is the most hilariously scary part of the recommendations:
Yes, nobody will have a problem with THAT clause.
The job of protecting against a systemic massive 9/11-style attack from a nation state in the cyber domain is rightfully the federal government's. But you can't replace a robust and realistic policy program with a Flood Insurance for Cyber. When Keith Alexander went around asking banks to give him access to their incoming traffic with a black box, they all said no, and for good reasons. Rob argues that not only should we go further than a black box doing network inspection, but this should apply to every company. It's a massive power grab and, luckily for all of us, a non-starter.

Remember, when Rob says this will encourage the adoption of best practices, what he means is "We are going to mandate how you run your networks, even though we cannot secure our own."

Monday, November 28, 2016

Unboxing "0day" for Policy People

Sometimes the bugs come out of the box.

Today's painful realization is that the very term "0day" has put this weird box around the policy brain, and minimized the dangers of regulation on all research in the security space, especially, for some reason, the European policy brain (including our British friends!). So I want to demonstrate some Zen Koans to help unbox you, so when Microsoft says they're looking "widely" with their "bounty" program you know what they mean.

Some things which are 0day, but outside the box:

  • Techniques for undetectable persistence on Windows 10
  • Ways to manipulate a heap on iOS that guarantee a certain heap layout
  • A function pointer that is always at a static location in Google Chrome and is called periodically
  • A way to send a lot of data using DNS through Microsoft Exchange servers
  • A shellcode that does something useful on Cisco's OS
  • Ways to clean up a process so that it continues nicely after exploitation.

If you think "Oh, they have promised not to regulate knowledge in general, just dangerous exploits!" then think again. There are many clauses in the Wassenaar agreements and every other proposed regulation (looking in Ari Schwartz's direction here) that seek to control exactly these things. Hopefully this post helps clarify why every security researcher had a big freakout with the Wassenaar proposals.

Wednesday, November 23, 2016

CIS VEP Panel Commentary

You can be super smart and not understand CNO operational issues because of a lack of experience in the area. And you can be smart and have ethical issues with the very idea of doing CNO. Above is a link to the CIS panel released last week on "Government Hacking" that discusses the VEP where both are on display.

It's hard to address the "ethical" issues around SIGINT collection that make people unhappy. I find it disturbing (as should you) that Ari Schwartz and Rob Knake and the Obama White House decided to do what they did with the VEP, sacrificing years of effort to maintain operational advantage by our IC, because of vague ethical issues with something they don't even understand fully. In the video, you can see Ari's face panic when the question comes in about what a vulnerability "Class" is, something we've written about on this blog. Sinan Eren answers it, much to Ari's relief, because Ari has no idea what a vulnerability class is except in the most general sense. He couldn't name them if his life depended on it. AND LIVES DO DEPEND ON IT.

It's also funny in the video to see Ari's look of surprise when he hears Sinan say "Vulnerabilities don't matter from a defensive perspective - focusing on mitigating factors is what makes the difference from a software security perspective". You can see an epiphany almost start to form in his head, then fade away as he returns to his blind ideology.

Inexperience with operational matters is something we can point out clearly though. You can always tell someone is inexperienced when they say things like "How long should we hold a vulnerability for?" or "You don't even need 0days to attack things!" That second one is true, except against hard targets, or when you cannot afford to get caught. Does that sound like the exact position the IC is in? Yes, yes it does.

This is the probabilistic game every good operator has in their heads. This is why it's not simple. Like a scuba operator measuring their outgassing a good CNO OPSEC person is also measuring their exposure to other operations across their entire toolchain at all times.
The reason hackers love 0day is not always the high success levels. It's the protection against detection by intermediaries or the target themselves. Likewise, it takes a very long time - sometimes years - to properly test an exploit in the wild. When people say "How long have you had this bug?" the answer from a properly trained operator is always "Not long enough to be comfortable with it".

The saddest part of the VEP video was when Ari says "Just because we've given it to a vendor doesn't mean it's blown!" Everyone in the IC was headslapping as he said that. It demonstrates a complete lack of understanding of how operations are protected that should not be the case in someone making policy that affects the IC.

But it comes out, during the video, that Ari believes we should control the whole vulnerability "market". That was his real goal with the VEP. And that means everyone. It means Ari thinks the entire research community should follow some disclosure law he and his friends think up and ram through Congress, without any understanding of the impact of his "Ethics" on the rest of us. It's the same as the Wassenaar Agreement. And yet the EFF is still trying to support him on this one. And that too, is baffling.

Sunday, November 20, 2016

you have to love it

No matter how many years it's been since I left - two decades now - people still look only at this one thing on my resume.

I want to spend a couple minutes pointing out the massive cultural differences between hackers and everyone else in this industry. Because as I crawl around in the policy world I sense these wide gulfs. The first one I sense is about books. Because hackers, despite coming from many backgrounds, share a common philosophy learned entirely from D&D, Neuromancer, Asimov, Heinlein, Snow Crash, Cryptonomicon, The Long Run, Dune, and all the other hard core science fiction they nestled in as they began the ascent up the mountain towards internalizing a difficult discipline.

Because of this, they are universally paranoid, atheistic, libertarian. And when the policy world tries to define norms without understanding the built-in philosophy of the domain, they run into fierce resistance from the denizens of this space, a lot of it only understandable if they have read the curriculum. I was astonished when Sue and other people I know in the legal/policy sphere hadn't read Snow Crash - I assumed EVERYONE had read it, because in my world, everyone has.

There are apparently moves to fire Rogers as DIRNSA. "Huge if true", as they say. I can't even remember any hints of this kind of action ever happening before.

NSA is far too important to this country to ruin, a shining jewel of technological prowess. We should be proud of it. It has gone through some hard times, most of which were not the fault of the organization. VEP is an example of our White House being ashamed of the NSA's competence. How can you expect to keep people if you are ashamed of them?

That then, is all we should ask as the American people of our next DIRNSA. You don't have to be a geek, or have read any science fiction. But it doesn't hurt. You have to join the culture you are becoming a part of. You cannot lead the NSA without falling in love with it.

And if you are in the cyber policy space as a lawyer, make the effort to read the cannon of cyber war, and we hackers will read your books about more recent history.

Many of these are free. :)

<live list follows>

  1. The Long Run PDF 
  2. Snow Crash AMZN
  3. Cryptonomicon AMZN
  4. Dune AMZN

Saturday, November 19, 2016

The State of Cyber Norms

It's worth pointing out that despite the insanely optimistic musings (1,2,3) from everyone in the State Dept about the progress of the international relations world on cyber norms, the reality is a disaster.

The shining light, which State and the Obama administration completely get credit for is the dissolution of the Chinese State economic espionage strategy. But that ignores the overall picture:

  • The UN GGE process misses clear players (Russia/China) and has at the root the issue of nobody agreeing on any of the definitions of the words they use
  • NATO's Tallinn process has many key problems (i.e. it is largely disconnected from the realities of the domain's technical characteristics) 
  • The US's transparency about our SIGINT process has been met with nothing from its European partners, who continue to batter us with hypocritical cries about privacy post-Snowden
To put it in the clearest possible terms: Nobody at State had the foresight to delimit "Not messing with our election" to the  Russians, which meant we had to get into a last minute massive escalation game with them instead. In addition to the lack of progress on any realistic front from our more traditional international efforts, this is the kind of total failure that needs to be publicly recognized.

Here is an example paper exploring what collateral damage might mean. Imagine trying to apply international law or norms to a domain where you don't even know what collateral damage is yet! 

Tuesday, November 8, 2016

New Agencies We Need in the Next Administration

National Cyber Forensics Agency

We need an agency in charge of decryption of phones and analysis of data. It's not just about managing the decryption tools themselves, which are going to remain secret and not handed out to local PDs and FBI offices, but gaining the know-how of how you do forensics and data minimization in a robust way to protect US civil liberties.

This is going to cost a lot more money than I think people are expecting, but we have to do it, and the longer we wait, the more expensive it will be to bootstrap.

National Active Defense Agency

Marketing buzz has ruined the term "Active Defense". But "hack-back" is unpalatably honest. However, if you keep a careful eye on the policy groups, they are quickly finding ways to lay the groundwork for an agency that uses private dollars to hack back against Chinese/Russian C2, and legalize active measures against botnets and worms such as MIRAI.

This is not as hard legally and politically as people sometimes make it sound. You just run it like a penetration testing company, with scope and authority from DHS and money and talent from the private sector. And you make the State Dept sell it overseas, because that's their job and we work with the cyber norms we have, not the ones we want, sometimes.

National CISO

CISO is one of those jobs that destroys people. Thankless, and with the cloud of doom sticking to your pant legs like a toddler's poo everywhere you go. But we need, not centralization, but clarity of vision and of quality and, frankly, someone to give our executives in Government the straight dope of what they can and can't do with regards to their own IT infrastructure. We need a salesperson who can sell a unified government security fabric to all of the many business units that make up the Federal Government. So far we've concentrated on finding bureaucrats with authorities.

Every big bank has the identical federated business plan as the USG when it comes to how this sort of information security and IT infrastructure needs to be run. We need to copy their DNA and figure out how to do this, if not right, at least a lot less wrong.

Saturday, November 5, 2016

It's Raining VEP!

This is the technical community's feeling on these papers, summed up in a tweet.

And this is it summed up in...another tweet.

There's this weird trend in the policy world to have a group of law students get together, read a bunch of Wired and NYT articles on all things cyber, quote a few of their own papers on the subject and call it "research". That's like doing lemur research by looking at the drawings a first grade class does after they watch Madagascar. If they really wanted to do research on the 0day market, they should find some 0day, then sell them on the "market"! Can't? Then STFU about all things vuln markets!

But to get more substantive on a critique of the latest Pro-VEP paper (this time by @jason_healey and co). Let's start with the most ridiculous recommendation in any VEP paper so far!
Yes, because opportunities just wait around for the VEP process to complete.

Just to revisit how hacking works: your shock troops are small teams building and deploying exploits faster than their hard targets can adjust. Custom-built is what works best for both return on investment and OPSEC. Policies that apply to "every vulnerability" ignore the vast differences in how the Government finds and uses vulnerabilities across its entire mission set. Do we have an entire-Government-wide policy on shovels?

Aside from the data being of "who the fuck knows" in terms of accuracy, and the contents of the table being entirely conjecture by law students who've never touched a vulnerability, I can't imagine the the purchasing officer who is going to spend 4.4M USD on things they don't get to use. In what world do people think that's how the system works? SEQUESTRATION IS A REAL THING, THIS TABLE IS NOT.

Look, not for nothing, but you need lots more bugs than 45 to get the job done, and beyond that, to get the job done safely. Your instinct should tell you that a team with 8 Solaris implants (c.f. EQGRP latest leak) has more than 15 bugs a year? THINK ABOUT IT: 8 DIFFERENT SOLARIS IMPLANTS. IMAGINE THE REASONS WHY.

THIS is just a description of the ethical argument in favor of unilaterally deciding not to do SIGINT. Not one I think we should use when designing government policy.
Here's what it comes down to: Unrealistic expectations of some sort of ethical purity argument as applied to SIGINT policy. The icing on the cake is the below feeling of betrayal when reality steps in.

I guarantee you that Obama, who is involved in our offensive operations more than any previous President, does not intend what VEP supporters think he intended. The key is the phrase "or the technology sector expects". That's what this is about. The Tech Sector has expectations. They are not being met, nor should they be.

Friday, November 4, 2016

Rule 41 Panel Discussion

So the link above is to the Government Hacking: Rule 41 talk held on Nov 2, 2016 panel with Granick, Salgado, Cocker, and several other lawyers heavily involved in the space.

I want to sum this up with two clear things:

  1. Whether you think the rule 41 changes are Substantive (and hence should be delayed and addressed by Congress) or simply a procedural change and therefor should be passed immediately to address major gaps - appears to depend entirely upon who is paying you to argue which side, as Granick points out 
  2. The Government (pro Rule Change, particularly Anello at ~1:40) side appears to intentionally confuse the issue. On one hand they state that it doesn't matter that most of the computers they will be hacking under this order are international because that's handled by multilateral treaties and cannot be handled with Rules or even by Congress. They acknowledge that there's no current way to handle the international issue by law or to handle this issue in the current MLATs. Not only that, there's no way we'd be comfortable with any other country doing what we're proposing to do under this Rule. On the other hand Allison Bragg claims that this does not change that we need the rule - as if there is some way we can make "anonymous computers" definitively domestic pursuant to our searches in some way. 

Straight up, Allison Bragg claims that the Rule change says that "if you don't know where a computer is, if it is domestic, you'll be able to apply Rule 41 to it". That makes NO SENSE from a technical perspective. And it is something that should be called out a bit further by Granick, Cocker and the others on the panel.

It would not be incorrect to state that Bragg thinks "We need this, no matter what the issues with it are, because otherwise we can't prosecute anything". This is probably short-sighted, especially considering this Rule change is not limited to terrorism or child porn cases, but in fact, across the whole board of crimes.

There are a lot of other issues in this area that are not REALLY Rule 41 change-related (particularity concerns with the 4thA, etc.) but are slightly touched upon by the panel, which is full of super-strong legal minds fun to listen to. Although if you are reading this, you probably don't have the time. :)

Thursday, November 3, 2016

The Ethical Argument against VEP

I promised myself I wouldn't write about VEP anymore on this blog. But we have reached stage 2 in the argument and it's important to note that originally everyone was claiming the VEP was necessary as a function of public safety, to coordinate defense against systemic risks that were pertinent to our national structure. Now they say it is an ethical issue, as Ari eventually did on stage at CyCon.

More correctly, the idea of the Government holding exploits, and in particular the NSA, makes people feel "icky" and when you talk to congressional staffers they don't trust the NSA to make decisions in the best interests of the American people with the exploits they do have and use. People who support the VEP rationalize their feelings of ickyness and distrust into an "Ethics problem".

Mozilla and Microsoft and Google and every other large software company would love to make it seem like the Ethics of the issue basically requires that the government get out of the business of having and using exploits at all. But they don't secure their systems because of ethical issues - they secure them because of market forces.

From a purely ethical issue, who knows: it may be that SIGINT is an unethical thing to ever do. Or it may be that it is a proportional and reasonable response to our national security needs. If we want to get out of the SIGINT business, we should just say so.

To put this in concrete terms for Jeff: Going through with the VEP is eventually going to require strict export controls on what you are allowed to say at Defcon. The ethical judgments on that seem to point towards a less free society.

Wednesday, November 2, 2016

Quick note on Tallinn and International Law of Conflict issues in Cyber

Check out what the Marine Colonel Gary Brown says at minute 28 of this CSIS video from last year. Some of the best explanation of Tallinn's flaws I've ever heard. Way better than how I say it. :)

One interesting thing that I don't know is addressed in International Law is that cyber splits up whether a State is responsible and WHICH State is responsible.

Tuesday, November 1, 2016

The GWU Active Defense Report is a Secret Argument for Cyber Letters of Marque!

Let's talk about the Active Defense Report from GWU, starting with who wrote it, but not to bury the leed, this paper is all about an argument FOR CYBER LETTERS OF MARQUE! :)

It's always a bad sign when you have Jane Holl Lute talking about anything  computer related. She spent her entire BlackHat talk saying how little she knew about the subject. This tells me they picked people by who had titles, not knowledge. But they also have Tim Evans and Nate Fick and Stewart Baker and other people who clearly DO have experience in the subject. 
The first thing the paper does is of course to try to define active defense. If we go over the history of the term, it was synonymous for hacking back, and occasionally for doing some other pretty obviously illegal things, until the marketing droids at Crowdstrike who were using it daily got tons of heat and had to walk it backwards about ten steps by including a lot of stuff that is not at all active defense in their marketing material and public statements.

This paper is no different.

On the left, stuff that is in no way related to active defense! On the right, stuff that is clearly illegal! For example, "Hunting" is "looking at logs to find patterns". Let's not fall for our own marketing BS.
The smart thing to do with this paper is ignore entirely the Executive Summary where they suggest about a million things for the government to do, including the executive branch, congress, various government agencies, etc. This would be a tremendous effort of ungodly proportions! And I think it occludes the important nature of this paper, which is buried far below.

This section is where the paper starts to acknowledge the issues at hand: Government is failing to provide a protective security umbrella.

This paper drives relentlessly towards one conclusion: We need a new model for Cyber Letters of Marque. It starts subtly.

Note the highlighted section that we call "Foreshadowing".

The paper goes a bit more explicit with why this is important in the next few sections.

Let's talk about that example, because it is clearly "Hack Back", the real and only definition of active defense. But the Government is acknowledging that it currently uses selective prosecution as its current plan for Cyber Letters of Marque, and needs a new, more explicit model, which it goes on to explain as essentially the exact thing we have in this blog post over here.

The snippet from the Appendix (written by the Center for Democracy and Technology) shows how most people in Government feel about these ideas (uneasy), but they will have to get used to it.

What should make us MORE uneasy, as the paper presents in the conclusion, is the idea of doing nothing, and allowing the norm of "Not acting unless we feel like it" to be the international rule of the road. The reason cyber letters of marque are a good idea is that they explicitly address what is clearly already happening, while allowing for resources from private companies to directly solve the problems private companies are having.

While private companies are not going to be directing hacking against C2 infrastructure in this model, they will be paying for it, and getting their priorities met. This addresses a significant gap in nation-state sovereignty, and the authors of the paper argue that it needs to happen as soon as possible. This paper sneaked these ideas in there in diplomatic terms, which is a very interesting development, to say the least.