Page 99

Hacking the Tax Code

The tax code isn’t software. It doesn’t run on a computer. But it’s still code. It’s a series of algorithms that takes an input—financial information for the year—and produces an output: the amount of tax owed. It’s incredibly complex code; there are a bazillion details and exceptions and special cases. It consists of government laws, rulings from the tax authorities, judicial decisions, and legal opinions.

Like computer code, the tax code has bugs. They might be mistakes in how the tax laws were written. They might be mistakes in how the tax code is interpreted, oversights in how parts of the law were conceived, or unintended omissions of some sort or another. They might arise from the exponentially huge number of ways different parts of the tax code interact.

A recent example comes from the 2017 Tax Cuts and Jobs Act. That law was drafted in both haste and secret, and quickly passed without any time for review—or even proofreading. One of the things in it was a typo that accidentally categorized military death benefits as earned income. The practical effect of that mistake is that surviving family members were hit with surprise tax bills of US$10,000 or more.

That’s a bug, but not a vulnerability. An example of a vulnerability is the “Double Irish with a Dutch Sandwich.” It arises from the interactions of tax laws in multiple countries, and it’s how companies like Google and Apple have avoided paying U.S. taxes despite being U.S. companies. Estimates are that U.S. companies avoided paying nearly US$200 billion in taxes in 2017 alone.

In the tax world, vulnerabilities are called loopholes. Exploits are called tax avoidance strategies. And there are thousands of black-hat researchers who examine every line of the tax code looking for exploitable vulnerabilities—tax attorneys and tax accountants.

Some vulnerabilities are deliberately created. Lobbyists are constantly trying to insert this or that provision into the tax code that benefits their clients financially. That same 2017 U.S. tax law included a special tax break for oil and gas investment partnerships, a special exemption that ensures that fewer than 1 in 1,000 estates will have to pay estate tax, and language specifically expanding a pass-through loophole that industry uses to incorporate companies offshore and avoid U.S. taxes. That’s not hacking the tax code. It’s hacking the processes that create them: the legislative process that creates tax law.

We know the processes to use to fix vulnerabilities in computer code. Before the code is finished, we can employ some sort of secure development processes, with automatic bug-finding tools and maybe source code audits. After the code is deployed, we might rely on vulnerability finding by the security community, perhaps bug bounties—and most of all, quick patching when vulnerabilities are discovered.

What does it mean to “patch” the tax code? Passing any tax legislation is a big deal, especially in the United States where the issue is so partisan and contentious. (That 2017 earned income tax bug for military families hasn’t yet been fixed. And that’s an easy one; everyone acknowledges it was a mistake.) We don’t have the ability to patch tax code with anywhere near the same agility that we have to patch software.

We can patch some vulnerabilities, though. The other way tax code is modified is by IRS and judicial rulings. The 2017 tax law capped income tax deductions for property taxes. This provision didn’t come into force in 2018, so someone came up with the clever hack to prepay 2018 property taxes in 2017. Just before the end of the year, the IRS ruled about when that was legal and when it wasn’t. Short answer: most of the time, it wasn’t.

There’s another option: that the vulnerability isn’t patched and isn’t explicitly approved, and slowly becomes part of the normal way of doing things. Lots of tax loopholes end up like this. Sometimes they’re even given retroactive legality by the IRS or Congress after a constituency and lobbying effort gets behind them. This process is how systems evolve. A hack subverts the intent of a system. Whatever governing system has jurisdiction either blocks the hack or allows it—or does nothing and the hack becomes the new normal.

Here’s my question: what happens when artificial intelligence and machine learning (ML) gets hold of this problem? We already have ML systems that find software vulnerabilities. What happens when you feed a ML system the entire U.S. tax code and tell it to figure out all of the ways to minimize the amount of tax owed? Or, in the case of a multinational corporation, to feed it the entire planet’s tax codes? What sort of vulnerabilities would it find? And how many? Dozens or millions?

In 2015, Volkswagen was caught cheating on emissions control tests. It didn’t forge test results; it got the cars’ computers to cheat for them. Engineers programmed the software in the car’s onboard computer to detect when the car was undergoing an emissions test. The computer then activated the car’s emissions-curbing systems, but only for the duration of the test. The result was that the cars had much better performance on the road at the cost of producing more pollution.

ML will result in lots of hacks like this. They’ll be more subtle. They’ll be even harder to discover. It’s because of the way ML systems optimize themselves, and because their specific optimizations can be impossible for us humans to understand. Their human programmers won’t even know what’s going on.

Any good ML system will naturally find and exploit hacks. This is because their only constraints are the rules of the system. If there are problems, inconsistencies, or loopholes in the rules, and if those properties lead to a “better” solution as defined by the program, then those systems will find them. The challenge is that you have to define the system’s goals completely and precisely, and that that’s impossible.

The tax code can be hacked. Financial markets regulations can be hacked. The market economy, democracy itself, and our cognitive systems can all be hacked. Tasking a ML system to find new hacks against any of these is still science fiction, but it’s not stupid science fiction. And ML will drastically change how we need to think about policy, law, and government. Now’s the time to figure out how.

This essay originally appeared in the September/October 2020 issue of IEEE Security & Privacy. I wrote it when I started writing my latest book, but never published it here.

Tags: , , , , ,

Posted on February 10, 2023 at 6:24 AM • View Comments

Mary Queen of Scots Letters Decrypted

This is a neat piece of historical research.

The team of computer scientist George Lasry, pianist Norbert Biermann and astrophysicist Satoshi Tomokiyo—all keen cryptographers—initially thought the batch of encoded documents related to Italy, because that was how they were filed at the Bibliothèque Nationale de France.

However, they quickly realised the letters were in French. Many verb and adjectival forms being feminine, regular mention of captivity, and recurring names—such as Walsingham—all put them on the trail of Mary. Sir Francis Walsingham was Queen Elizabeth’s spymaster.

The code was a simple replacement system in which symbols stand either for letters, or for common words and names. But it would still have taken centuries to crunch all the possibilities, so the team used an algorithm that homed in on likely solutions.

Academic paper.

EDITED TO ADD (2/13): More news.

Tags: , ,

Posted on February 9, 2023 at 7:15 AM • View Comments

SolarWinds and Market Incentives

In early 2021, IEEE Security and Privacy asked a number of board members for brief perspectives on the SolarWinds incident while it was still breaking news. This was my response.

The penetration of government and corporate networks worldwide is the result of inadequate cyberdefenses across the board. The lessons are many, but I want to focus on one important one we’ve learned: the software that’s managing our critical networks isn’t secure, and that’s because the market doesn’t reward that security.

SolarWinds is a perfect example. The company was the initial infection vector for much of the operation. Its trusted position inside so many critical networks made it a perfect target for a supply-chain attack, and its shoddy security practices made it an easy target.

Why did SolarWinds have such bad security? The answer is because it was more profitable. The company is owned by Thoma Bravo partners, a private-equity firm known for radical cost-cutting in the name of short-term profit. Under CEO Kevin Thompson, the company underspent on security even as it outsourced software development. The New York Times reports that the company’s cybersecurity advisor quit after his “basic recommendations were ignored.” In a very real sense, SolarWinds profited because it secretly shifted a whole bunch of risk to its customers: the US government, IT companies, and others.

This problem isn’t new, and, while it’s exacerbated by the private-equity funding model, it’s not unique to it. In general, the market doesn’t reward safety and security—especially when the effects of ignoring those things are long term and diffuse. The market rewards short-term profits at the expense of safety and security. (Watch and see whether SolarWinds suffers any long-term effects from this hack, or whether Thoma Bravo’s bet that it could profit by selling an insecure product was a good one.)

The solution here is twofold. The first is to improve government software procurement. Software is now critical to national security. Any system of procuring that software needs to evaluate the security of the software and the security practices of the company, in detail, to ensure that they are sufficient to meet the security needs of the network they’re being installed in. If these evaluations are made public, along with the list of companies that meet them, all network buyers can benefit from them. It’s a win for everybody.

But that isn’t enough; we need a second part. The only way to force companies to provide safety and security features for customers is through regulation. This is true whether we want seat belts in our cars, basic food safety at our restaurants, pajamas that don’t catch on fire, or home routers that aren’t vulnerable to cyberattack. The government needs to set minimum security standards for software that’s used in critical network applications, just as it sets software standards for avionics.

Without these two measures, it’s just too easy for companies to act like SolarWinds: save money by skimping on safety and security and hope for the best in the long term. That’s the rational thing for companies to do in an unregulated market, and the only way to change that is to change the economic incentives.

This essay originally appeared in the March/April 2021 issue of IEEE Security & Privacy.” I forgot to publish it here.

Tags: , , ,

Posted on February 8, 2023 at 6:46 AM • View Comments

Malware Delivered through Google Search

Criminals using Google search ads to deliver malware isn’t new, but Ars Technica declared that the problem has become much worse recently.

The surge is coming from numerous malware families, including AuroraStealer, IcedID, Meta Stealer, RedLine Stealer, Vidar, Formbook, and XLoader. In the past, these families typically relied on phishing and malicious spam that attached Microsoft Word documents with booby-trapped macros. Over the past month, Google Ads has become the go-to place for criminals to spread their malicious wares that are disguised as legitimate downloads by impersonating brands such as Adobe Reader, Gimp, Microsoft Teams, OBS, Slack, Tor, and Thunderbird.

[…]

It’s clear that despite all the progress Google has made filtering malicious sites out of returned ads and search results over the past couple decades, criminals have found ways to strike back. These criminals excel at finding the latest techniques to counter the filtering. As soon as Google devises a way to block them, the criminals figure out new ways to circumvent those protections.

Tags: , ,

Posted on February 7, 2023 at 7:23 AM • View Comments

Attacking Machine Learning Systems

The field of machine learning (ML) security—and corresponding adversarial ML—is rapidly advancing as researchers develop sophisticated techniques to perturb, disrupt, or steal the ML model or data. It’s a heady time; because we know so little about the security of these systems, there are many opportunities for new researchers to publish in this field. In many ways, this circumstance reminds me of the cryptanalysis field in the 1990. And there is a lesson in that similarity: the complex mathematical attacks make for good academic papers, but we mustn’t lose sight of the fact that insecure software will be the likely attack vector for most ML systems.

We are amazed by real-world demonstrations of adversarial attacks on ML systems, such as a 3D-printed object that looks like a turtle but is recognized (from any orientation) by the ML system as a gun. Or adding a few stickers that look like smudges to a stop sign so that it is recognized by a state-of-the-art system as a 45 mi/h speed limit sign. But what if, instead, somebody hacked into the system and just switched the labels for “gun” and “turtle” or swapped “stop” and “45 mi/h”? Systems can only match images with human-provided labels, so the software would never notice the switch. That is far easier and will remain a problem even if systems are developed that are robust to those adversarial attacks.

At their core, modern ML systems have complex mathematical models that use training data to become competent at a task. And while there are new risks inherent in the ML model, all of that complexity still runs in software. Training data are still stored in memory somewhere. And all of that is on a computer, on a network, and attached to the Internet. Like everything else, these systems will be hacked through vulnerabilities in those more conventional parts of the system.

This shouldn’t come as a surprise to anyone who has been working with Internet security. Cryptography has similar vulnerabilities. There is a robust field of cryptanalysis: the mathematics of code breaking. Over the last few decades, we in the academic world have developed a variety of cryptanalytic techniques. We have broken ciphers we previously thought secure. This research has, in turn, informed the design of cryptographic algorithms. The classified world of the NSA and its foreign counterparts have been doing the same thing for far longer. But aside from some special cases and unique circumstances, that’s not how encryption systems are exploited in practice. Outside of academic papers, cryptosystems are largely bypassed because everything around the cryptography is much less secure.

I wrote this in my book, Data and Goliath:

The problem is that encryption is just a bunch of math, and math has no agency. To turn that encryption math into something that can actually provide some security for you, it has to be written in computer code. And that code needs to run on a computer: one with hardware, an operating system, and other software. And that computer needs to be operated by a person and be on a network. All of those things will invariably introduce vulnerabilities that undermine the perfection of the mathematics…

This remains true even for pretty weak cryptography. It is much easier to find an exploitable software vulnerability than it is to find a cryptographic weakness. Even cryptographic algorithms that we in the academic community regard as “broken”—meaning there are attacks that are more efficient than brute force—are usable in the real world because the difficulty of breaking the mathematics repeatedly and at scale is much greater than the difficulty of breaking the computer system that the math is running on.

ML systems are similar. Systems that are vulnerable to model stealing through the careful construction of queries are more vulnerable to model stealing by hacking into the computers they’re stored in. Systems that are vulnerable to model inversion—this is where attackers recover the training data through carefully constructed queries—are much more vulnerable to attacks that take advantage of unpatched vulnerabilities.

But while security is only as strong as the weakest link, this doesn’t mean we can ignore either cryptography or ML security. Here, our experience with cryptography can serve as a guide. Cryptographic attacks have different characteristics than software and network attacks, something largely shared with ML attacks. Cryptographic attacks can be passive. That is, attackers who can recover the plaintext from nothing other than the ciphertext can eavesdrop on the communications channel, collect all of the encrypted traffic, and decrypt it on their own systems at their own pace, perhaps in a giant server farm in Utah. This is bulk surveillance and can easily operate on this massive scale.

On the other hand, computer hacking has to be conducted one target computer at a time. Sure, you can develop tools that can be used again and again. But you still need the time and expertise to deploy those tools against your targets, and you have to do so individually. This means that any attacker has to prioritize. So while the NSA has the expertise necessary to hack into everyone’s computer, it doesn’t have the budget to do so. Most of us are simply too low on its priorities list to ever get hacked. And that’s the real point of strong cryptography: it forces attackers like the NSA to prioritize.

This analogy only goes so far. ML is not anywhere near as mathematically sound as cryptography. Right now, it is a sloppy misunderstood mess: hack after hack, kludge after kludge, built on top of each other with some data dependency thrown in. Directly attacking an ML system with a model inversion attack or a perturbation attack isn’t as passive as eavesdropping on an encrypted communications channel, but it’s using the ML system as intended, albeit for unintended purposes. It’s much safer than actively hacking the network and the computer that the ML system is running on. And while it doesn’t scale as well as cryptanalytic attacks can—and there likely will be a far greater variety of ML systems than encryption algorithms—it has the potential to scale better than one-at-a-time computer hacking does. So here again, good ML security denies attackers all of those attack vectors.

We’re still in the early days of studying ML security, and we don’t yet know the contours of ML security techniques. There are really smart people working on this and making impressive progress, and it’ll be years before we fully understand it. Attacks come easy, and defensive techniques are regularly broken soon after they’re made public. It was the same with cryptography in the 1990s, but eventually the science settled down as people better understood the interplay between attack and defense. So while Google, Amazon, Microsoft, and Tesla have all faced adversarial ML attacks on their production systems in the last three years, that’s not going to be the norm going forward.

All of this also means that our security for ML systems depends largely on the same conventional computer security techniques we’ve been using for decades. This includes writing vulnerability-free software, designing user interfaces that help resist social engineering, and building computer networks that aren’t full of holes. It’s the same risk-mitigation techniques that we’ve been living with for decades. That we’re still mediocre at it is cause for concern, with regard to both ML systems and computing in general.

I love cryptography and cryptanalysis. I love the elegance of the mathematics and the thrill of discovering a flaw—or even of reading and understanding a flaw that someone else discovered—in the mathematics. It feels like security in its purest form. Similarly, I am starting to love adversarial ML and ML security, and its tricks and techniques, for the same reasons.

I am not advocating that we stop developing new adversarial ML attacks. It teaches us about the systems being attacked and how they actually work. They are, in a sense, mechanisms for algorithmic understandability. Building secure ML systems is important research and something we in the security community should continue to do.

There is no such thing as a pure ML system. Every ML system is a hybrid of ML software and traditional software. And while ML systems bring new risks that we haven’t previously encountered, we need to recognize that the majority of attacks against these systems aren’t going to target the ML part. Security is only as strong as the weakest link. As bad as ML security is right now, it will improve as the science improves. And from then on, as in cryptography, the weakest link will be in the software surrounding the ML system.

This essay originally appeared in the May 2020 issue of IEEE Computer. I forgot to reprint it here.

Tags: , , , , ,

Posted on February 6, 2023 at 6:02 AM • View Comments

A Hacker’s Mind News

A Hacker’s Mind will be published on Tuesday.

I have done a written interview and a podcast interview about the book. It’s been chosen as a “February 2023 Must-Read Book” by the Next Big Idea Club. And an “Editor’s Pick”—whatever that means—on Amazon.

There have been three reviews so far. I am hoping for more. And maybe even a published excerpt or two.

Amazon and others will start shipping the book on Tuesday. If you ordered a signed copy from me, it is already in the mail.

If you can leave a review somewhere, I would appreciate it.

Tags: , , ,

Posted on February 3, 2023 at 3:03 PM • View Comments

Manipulating Weights in Face-Recognition AI Systems

Interesting research: “Facial Misrecognition Systems: Simple Weight Manipulations Force DNNs to Err Only on Specific Persons“:

Abstract: In this paper we describe how to plant novel types of backdoors in any facial recognition model based on the popular architecture of deep Siamese neural networks, by mathematically changing a small fraction of its weights (i.e., without using any additional training or optimization). These backdoors force the system to err only on specific persons which are preselected by the attacker. For example, we show how such a backdoored system can take any two images of a particular person and decide that they represent different persons (an anonymity attack), or take any two images of a particular pair of persons and decide that they represent the same person (a confusion attack), with almost no effect on the correctness of its decisions for other persons. Uniquely, we show that multiple backdoors can be independently installed by multiple attackers who may not be aware of each other’s existence with almost no interference.

We have experimentally verified the attacks on a FaceNet-based facial recognition system, which achieves SOTA accuracy on the standard LFW dataset of 99.35%. When we tried to individually anonymize ten celebrities, the network failed to recognize two of their images as being the same person in 96.97% to 98.29% of the time. When we tried to confuse between the extremely different looking Morgan Freeman and Scarlett Johansson, for example, their images were declared to be the same person in 91.51% of the time. For each type of backdoor, we sequentially installed multiple backdoors with minimal effect on the performance of each one (for example, anonymizing all ten celebrities on the same model reduced the success rate for each celebrity by no more than 0.91%). In all of our experiments, the benign accuracy of the network on other persons was degraded by no more than 0.48% (and in most cases, it remained above 99.30%).

It’s a weird attack. On the one hand, the attacker has access to the internals of the facial recognition system. On the other hand, this is a novel attack in that it manipulates internal weights to achieve a specific outcome. Given that we have no idea how those weights work, it’s an important result.

Tags: , ,

Posted on February 3, 2023 at 7:07 AM • View Comments

AIs as Computer Hackers

Hacker “Capture the Flag” has been a mainstay at hacker gatherings since the mid-1990s. It’s like the outdoor game, but played on computer networks. Teams of hackers defend their own computers while attacking other teams’. It’s a controlled setting for what computer hackers do in real life: finding and fixing vulnerabilities in their own systems and exploiting them in others’. It’s the software vulnerability lifecycle.

These days, dozens of teams from around the world compete in weekend-long marathon events held all over the world. People train for months. Winning is a big deal. If you’re into this sort of thing, it’s pretty much the most fun you can possibly have on the Internet without committing multiple felonies.

In 2016, DARPA ran a similarly styled event for artificial intelligence (AI). One hundred teams entered their systems into the Cyber Grand Challenge. After completing qualifying rounds, seven finalists competed at the DEFCON hacker convention in Las Vegas. The competition occurred in a specially designed test environment filled with custom software that had never been analyzed or tested. The AIs were given 10 hours to find vulnerabilities to exploit against the other AIs in the competition and to patch themselves against exploitation. A system called Mayhem, created by a team of Carnegie-Mellon computer security researchers, won. The researchers have since commercialized the technology, which is now busily defending networks for customers like the U.S. Department of Defense.

There was a traditional human–team capture-the-flag event at DEFCON that same year. Mayhem was invited to participate. It came in last overall, but it didn’t come in last in every category all of the time.

I figured it was only a matter of time. It would be the same story we’ve seen in so many other areas of AI: the games of chess and go, X-ray and disease diagnostics, writing fake news. AIs would improve every year because all of the core technologies are continually improving. Humans would largely stay the same because we remain humans even as our tools improve. Eventually, the AIs would routinely beat the humans. I guessed that it would take about a decade.

But now, five years later, I have no idea if that prediction is still on track. Inexplicably, DARPA never repeated the event. Research on the individual components of the software vulnerability lifecycle does continue. There’s an enormous amount of work being done on automatic vulnerability finding. Going through software code line by line is exactly the sort of tedious problem at which machine learning systems excel, if they can only be taught how to recognize a vulnerability. There is also work on automatic vulnerability exploitation and lots on automatic update and patching. Still, there is something uniquely powerful about a competition that puts all of the components together and tests them against others.

To see that in action, you have to go to China. Since 2017, China has held at least seven of these competitions—called Robot Hacking Games—many with multiple qualifying rounds. The first included one team each from the United States, Russia, and Ukraine. The rest have been Chinese only: teams from Chinese universities, teams from companies like Baidu and Tencent, teams from the military. Rules seem to vary. Sometimes human–AI hybrid teams compete.

Details of these events are few. They’re Chinese language only, which naturally limits what the West knows about them. I didn’t even know they existed until Dakota Cary, a research analyst at the Center for Security and Emerging Technology and a Chinese speaker, wrote a report about them a few months ago. And they’re increasingly hosted by the People’s Liberation Army, which presumably controls how much detail becomes public.

Some things we can infer. In 2016, none of the Cyber Grand Challenge teams used modern machine learning techniques. Certainly most of the Robot Hacking Games entrants are using them today. And the competitions encourage collaboration as well as competition between the teams. Presumably that accelerates advances in the field.

None of this is to say that real robot hackers are poised to attack us today, but I wish I could predict with some certainty when that day will come. In 2018, I wrote about how AI could change the attack/defense balance in cybersecurity. I said that it is impossible to know which side would benefit more but predicted that the technologies would benefit the defense more, at least in the short term. I wrote: “Defense is currently in a worse position than offense precisely because of the human components. Present-day attacks pit the relative advantages of computers and humans against the relative weaknesses of computers and humans. Computers moving into what are traditionally human areas will rebalance that equation.”

Unfortunately, it’s the People’s Liberation Army and not DARPA that will be the first to learn if I am right or wrong and how soon it matters.

This essay originally appeared in the January/February 2022 issue of IEEE Security & Privacy.

Tags: , , , , ,

Posted on February 2, 2023 at 6:59 AM • View Comments

Passwords Are Terrible (Surprising No One)

This is the result of a security audit:

More than a fifth of the passwords protecting network accounts at the US Department of the Interior—including Password1234, Password1234!, and ChangeItN0w!—were weak enough to be cracked using standard methods, a recently published security audit of the agency found.

[…]

The results weren’t encouraging. In all, the auditors cracked 18,174—or 21 percent—­of the 85,944 cryptographic hashes they tested; 288 of the affected accounts had elevated privileges, and 362 of them belonged to senior government employees. In the first 90 minutes of testing, auditors cracked the hashes for 16 percent of the department’s user accounts.

The audit uncovered another security weakness—the failure to consistently implement multi-factor authentication (MFA). The failure extended to 25—­or 89 percent—­of 28 high-value assets (HVAs), which, when breached, have the potential to severely impact agency operations.

Original story:

To make their point, the watchdog spent less than $15,000 on building a password-cracking rig—a setup of a high-performance computer or several chained together ­- with the computing power designed to take on complex mathematical tasks, like recovering hashed passwords. Within the first 90 minutes, the watchdog was able to recover nearly 14,000 employee passwords, or about 16% of all department accounts, including passwords like ‘Polar_bear65’ and ‘Nationalparks2014!’.

Tags: , ,

Posted on February 1, 2023 at 7:08 AM • View Comments

← Earlier EntriesLater Entries →

Sidebar photo of Bruce Schneier by Joe MacInnis.