Page 4 of 6
Renesys is reporting that Internet traffic is being manipulatively rerouted, presumably for eavesdropping purposes. The attacks exploit flaws in the Border Gateway Protocol (BGP). Ars Technica has a good article explaining the details.
The odds that the NSA is not doing this sort of thing are basically zero, but I’m sure that their activities are going to be harder to discover.
The online anonymity network Tor is a high-priority target for the National Security Agency. The work of attacking Tor is done by the NSA‘s application vulnerabilities branch, which is part of the systems intelligence directorate, or SID. The majority of NSA employees work in SID, which is tasked with collecting data from communications systems around the world.
According to a top-secret NSA presentation provided by the whistleblower Edward Snowden, one successful technique the NSA has developed involves exploiting the Tor browser bundle, a collection of programs designed to make it easy for people to install and use the software. The trick identifies Tor users on the Internet and then executes an attack against their Firefox web browser.
The NSA refers to these capabilities as CNE, or computer network exploitation.
The first step of this process is finding Tor users. To accomplish this, the NSA relies on its vast capability to monitor large parts of the Internet. This is done via the agency’s partnership with US telecoms firms under programs codenamed Stormbrew, Fairview, Oakstar and Blarney.
The NSA creates “fingerprints” that detect HTTP requests from the Tor network to particular servers. These fingerprints are loaded into NSA database systems like XKeyscore, a bespoke collection and analysis tool that NSA boasts allows its analysts to see “almost everything” a target does on the Internet.
Using powerful data analysis tools with codenames such as Turbulence, Turmoil and Tumult, the NSA automatically sifts through the enormous amount of Internet traffic that it sees, looking for Tor connections.
Last month, Brazilian TV news show Fantastico showed screenshots of an NSA tool that had the ability to identify Tor users by monitoring Internet traffic.
The very feature that makes Tor a powerful anonymity service, and the fact that all Tor users look alike on the Internet, makes it easy to differentiate Tor users from other web users. On the other hand, the anonymity provided by Tor makes it impossible for the NSA to know who the user is, or whether or not the user is in the US.
After identifying an individual Tor user on the Internet, the NSA uses its network of secret Internet servers to redirect those users to another set of secret Internet servers, with the codename FoxAcid, to infect the user’s computer. FoxAcid is an NSA system designed to act as a matchmaker between potential targets and attacks developed by the NSA, giving the agency opportunity to launch prepared attacks against their systems.
Once the computer is successfully attacked, it secretly calls back to a FoxAcid server, which then performs additional attacks on the target computer to ensure that it remains compromised long-term, and continues to provide eavesdropping information back to the NSA.
Exploiting the Tor browser bundle
Tor is a well-designed and robust anonymity tool, and successfully attacking it is difficult. The NSA attacks we found individually target Tor users by exploiting vulnerabilities in their Firefox browsers, and not the Tor application directly.
This, too, is difficult. Tor users often turn off vulnerable services like scripts and Flash when using Tor, making it difficult to target those services. Even so, the NSA uses a series of native Firefox vulnerabilities to attack users of the Tor browser bundle.
According to the training presentation provided by Snowden, EgotisticalGiraffe exploits a type confusion vulnerability in E4X, which is an XML extension for JavaScript. This vulnerability exists in Firefox 11.0—16.0.2, as well as Firefox 10.0 ESR—the Firefox version used until recently in the Tor browser bundle. According to another document, the vulnerability exploited by EgotisticalGiraffe was inadvertently fixed when Mozilla removed the E4X library with the vulnerability, and when Tor added that Firefox version into the Tor browser bundle, but NSA were confident that they would be able to find a replacement Firefox exploit that worked against version 17.0 ESR.
The Quantum system
To trick targets into visiting a FoxAcid server, the NSA relies on its secret partnerships with US telecoms companies. As part of the Turmoil system, the NSA places secret servers, codenamed Quantum, at key places on the Internet backbone. This placement ensures that they can react faster than other websites can. By exploiting that speed difference, these servers can impersonate a visited website to the target before the legitimate website can respond, thereby tricking the target’s browser to visit a Foxacid server.
In the academic literature, these are called “man-in-the-middle” attacks, and have been known to the commercial and academic security communities. More specifically, they are examples of “man-on-the-side” attacks.
They are hard for any organization other than the NSA to reliably execute, because they require the attacker to have a privileged position on the Internet backbone, and exploit a “race condition” between the NSA server and the legitimate website. This top-secret NSA diagram, made public last month, shows a Quantum server impersonating Google in this type of attack.
The NSA uses these fast Quantum servers to execute a packet injection attack, which surreptitiously redirects the target to the FoxAcid server. An article in the German magazine Spiegel, based on additional top secret Snowden documents, mentions an NSA developed attack technology with the name of QuantumInsert that performs redirection attacks. Another top-secret Tor presentation provided by Snowden mentions QuantumCookie to force cookies onto target browsers, and another Quantum program to “degrade/deny/disrupt Tor access”.
This same technique is used by the Chinese government to block its citizens from reading censored Internet content, and has been hypothesized as a probable NSA attack technique.
The FoxAcid system
According to various top-secret documents provided by Snowden, FoxAcid is the NSA codename for what the NSA calls an “exploit orchestrator,” an Internet-enabled system capable of attacking target computers in a variety of different ways. It is a Windows 2003 computer configured with custom software and a series of Perl scripts. These servers are run by the NSA’s tailored access operations, or TAO, group. TAO is another subgroup of the systems intelligence directorate.
The servers are on the public Internet. They have normal-looking domain names, and can be visited by any browser from anywhere; ownership of those domains cannot be traced back to the NSA.
However, if a browser tries to visit a FoxAcid server with a special URL, called a FoxAcid tag, the server attempts to infect that browser, and then the computer, in an effort to take control of it. The NSA can trick browsers into using that URL using a variety of methods, including the race-condition attack mentioned above and frame injection attacks.
FoxAcid tags are designed to look innocuous, so that anyone who sees them would not be suspicious. http://baseball2.2ndhalfplays.com/nested/attribs/bins/1/define/forms9952_z1zzz.html is an example of one such tag, given in another top-secret training presentation provided by Snowden.
There is no currently registered domain name by that name; it is just an example for internal NSA training purposes.
The training material states that merely trying to visit the homepage of a real FoxAcid server will not result in any attack, and that a specialized URL is required. This URL would be created by TAO for a specific NSA operation, and unique to that operation and target. This allows the FoxAcid server to know exactly who the target is when his computer contacts it.
According to Snowden, FoxAcid is a general CNE system, used for many types of attacks other than the Tor attacks described here. It is designed to be modular, with flexibility that allows TAO to swap and replace exploits if they are discovered, and only run certain exploits against certain types of targets.
The most valuable exploits are saved for the most important targets. Low-value exploits are run against technically sophisticated targets where the chance of detection is high. TAO maintains a library of exploits, each based on a different vulnerability in a system. Different exploits are authorized against different targets, depending on the value of the target, the target’s technical sophistication, the value of the exploit, and other considerations.
In the case of Tor users, FoxAcid might use EgotisticalGiraffe against their Firefox browsers.
According to a top-secret operational management procedures manual provided by Snowden, once a target is successfully exploited it is infected with one of several payloads. Two basic payloads mentioned in the manual are designed to collect configuration and location information from the target computer so an analyst can determine how to further infect the computer.
These decisions are made in part by the technical sophistication of the target and the security software installed on the target computer, called Personal Security Products or PSP, in the manual.
FoxAcid payloads are updated regularly by TAO. For example, the manual refers to version 8.2.1.1 of one of them.
FoxAcid servers also have sophisticated capabilities to avoid detection and to ensure successful infection of its targets. The operations manual states that a FoxAcid payload with the codename DireScallop can circumvent commercial products that prevent malicious software from making changes to a system that survive a reboot process.
The NSA also uses phishing attacks to induce users to click on FoxAcid tags.
TAO additionally uses FoxAcid to exploit callbacks—which is the general term for a computer infected by some automatic means—calling back to the NSA for more instructions and possibly to upload data from the target computer.
According to a top-secret operational management procedures manual, FoxAcid servers configured to receive callbacks are codenamed FrugalShot. After a callback, the FoxAcid server may run more exploits to ensure that the target computer remains compromised long term, as well as install “implants” designed to exfiltrate data.
By 2008, the NSA was getting so much FoxAcid callback data that they needed to build a special system to manage it all.
This essay previously appeared in the Guardian. It is the technical article associated with this more general-interest article. I also wrote two commentaries on the material.
EDITED TO ADD: Here is the source material we published. The Washington Post published its own story independently, based on some of the same source material and some new source material.
Here’s the official US government response to the story.
The Guardian decided to change the capitalization of the NSA codenames. They should properly be in all caps: FOXACID, QUANTUMCOOKIE, EGOTISTICALGIRAFFE, TURMOIL, and so on.
This is the relevant quote from the Spiegel article:
According to the slides in the GCHQ presentation, the attack was directed at several Belgacom employees and involved the planting of a highly developed attack technology referred to as a “Quantum Insert” (“QI”). It appears to be a method with which the person being targeted, without their knowledge, is redirected to websites that then plant malware on their computers that can then manipulate them. Some of the employees whose computers were infiltrated had “good access” to important parts of Belgacom’s infrastructure, and this seemed to please the British spies, according to the slides.
That should be “QUANTUMINSERT.” This is getting frustrating. The NSA really should release a style guide for press organizations publishing their secrets.
And the URL in the essay (now redacted at the Guardian site) was registered within minutes of the story posting, and is being used to serve malware. Don’t click on it.
The Brazilian television show “Fantastico” exposed an NSA training presentation that discusses how the agency runs man-in-the-middle attacks on the Internet. The point of the story was that the NSA engages in economic espionage against Petrobras, the Brazilian giant oil company, but I’m more interested in the tactical details.
The video on the webpage is long, and includes what I assume is a dramatization of an NSA classroom, but a few screen shots are important. The pages from the training presentation describe how the NSA’s MITM attack works:
However, in some cases GCHQ and the NSA appear to have taken a more aggressive and controversial route—on at least one occasion bypassing the need to approach Google directly by performing a man-in-the-middle attack to impersonate Google security certificates. One document published by Fantastico, apparently taken from an NSA presentation that also contains some GCHQ slides, describes “how the attack was done” to apparently snoop on SSL traffic. The document illustrates with a diagram how one of the agencies appears to have hacked into a target’s Internet router and covertly redirected targeted Google traffic using a fake security certificate so it could intercept the information in unencrypted format.
Documents from GCHQ’s “network exploitation” unit show that it operates a program called “FLYING PIG” that was started up in response to an increasing use of SSL encryption by email providers like Yahoo, Google, and Hotmail. The FLYING PIG system appears to allow it to identify information related to use of the anonymity browser Tor (it has the option to query “Tor events“) and also allows spies to collect information about specific SSL encryption certificates.
It’s that first link—also here—that shows the MITM attack against Google and its users.
Another screenshot implies is that the 2011 DigiNotar hack was either the work of the NSA, or exploited by the NSA.
Here’s another story on this.
Last week, a story broke about how Nokia mounts man-in-the-middle attacks against secure browser sessions.
The Finnish phone giant has since admitted that it decrypts secure data that passes through HTTPS connections—including social networking accounts, online banking, email and other secure sessions—in order to compress the data and speed up the loading of Web pages.
The basic problem is that https sessions are opaque as they travel through the network. That’s the point—it’s more secure—but it also means that the network can’t do anything about them. They can’t be compressed, cached, or otherwise optimized. They can’t be rendered remotely. They can’t be inspected for security vulnerabilities. All the network can do is transmit the data back and forth.
But in our cloud-centric world, it makes more and more sense to process web data in the cloud. Nokia isn’t alone here. Opera’s mobile browser performs all sorts of optimizations on web pages before they are sent over the air to your smart phone. Amazon does the same thing with browsing on the Kindle. MobileScope, a really good smart-phone security application, performs the same sort of man-in-the-middle attack against https sessions to detect and prevent data leakage. I think Umbrella does as well. Nokia’s mistake was that they did it without telling anyone. With appropriate consent, it’s perfectly reasonable for most people and organizations to give both performance and security companies that ability to decrypt and re-encrypt https sessions—at least most of the time.
This is an area where security concerns are butting up against other issues. Nokia’s answer, which is basically “trust us, we’re not looking at your data,” is going to increasingly be the norm.
This sort of attack will become more common as banks require two-factor authentication:
Tatanga checks the user account details including the number of accounts, supported currency, balance/limit details. It then chooses the account from which it could steal the highest amount.
Next, it initiates a transfer.
At this point Tatanga uses a Web Inject to trick the user into believing that the bank is performing a chipTAN test. The fake instructions request that the user generate a TAN for the purpose of this “test” and enter the TAN.
Note that the attack relies on tricking the user, which isn’t very hard.
This essay uses the interesting metaphor of the man-in-the-middle attacker to describe cloud providers like Facebook and Google. Basically, they get in the middle of our interactions with others and eavesdrop on the data going back and forth.
In 2005, I wrote an essay called “The Failure of Two-Factor Authentication,” where I predicted that attackers would get around multi-factor authentication systems with tools that attack the transactions in real time: man-in-the-middle attacks and Trojan attacks against the client endpoint.
This BBC article describes exactly that:
After logging in to the bank’s real site, account holders are being tricked by the offer of training in a new “upgraded security system”.
Money is then moved out of the account but this is hidden from the user.
[…]
Called a Man in the Browser (MitB) attack, the malware lives in the web browser and can get between the user and the website, altering what is seen and changing details of what is being entered.
The solution is to authenticate the transaction, not the person.
EDITED TO ADD (2/6): Another link.
It’s the Browser Exploit Against SSL/TLS Tool, or BEAST:
The tool is based on a blockwise-adaptive chosen-plaintext attack, a man-in-the-middle approach that injects segments of plain text sent by the target’s browser into the encrypted request stream to determine the shared key. The code can be injected into the user’s browser through JavaScript associated with a malicious advertisement distributed through a Web ad service or an IFRAME in a linkjacked site, ad, or other scripted elements on a webpage.
Using the known text blocks, BEAST can then use information collected to decrypt the target’s AES-encrypted requests, including encrypted cookies, and then hijack the no-longer secure connection. That decryption happens slowly, however; BEAST currently needs sessions of at least a half-hour to break cookies using keys over 1,000 characters long.
The attack, according to Duong, is capable of intercepting sessions with PayPal and other services that still use TLS 1.0which would be most secure sites, since follow-on versions of TLS aren’t yet supported in most browsers or Web server implementations.
While Rizzo and Duong believe BEAST is the first attack against SSL 3.0 that decrypts HTTPS requests, the vulnerability that BEAST exploits is well-known; BT chief security technology officer Bruce Schneier and UC Berkeley’s David Wagner pointed out in a 1999 analysis of SSL 3.0 that “SSL will provide a lot of known plain-text to the eavesdropper, but there seems to be no better alternative.” And TLS’s vulnerability to man-in-the middle attacks was made public in 2009. The IETF’s TLS Working Group published a fix for the problem, but the fix is unsupported by SSL.
Another article.
EDITED TO ADD: Good analysis.
It’s an easy attack. Register a domain that’s like your target except for a typo. So it would be countrpane.com instead of counterpane.com, or mailcounterpane.com instead of mail.counterpane.com. Then, when someone mistypes an e-mail address to someone at that company and you receive it, just forward it on as if nothing happened.
These are called “doppelganger domains.”
To test the vulnerability, the researchers set up 30 doppelganger accounts for various firms and found that the accounts attracted 120,000 e-mails in the six-month testing period.
The e-mails they collected included one that listed the full configuration details for the external Cisco routers of a large IT consulting firm, along with passwords for accessing the devices. Another e-mail going to a company outside the U.S. that manages motorway toll systems provided information for obtaining full VPN access into the system that supports the road tollways. The e-mail included information about the VPN software, usernames, and passwords.
They’re already being used to spy on companies:
Some of the companies whose doppelganger domains have already been taken by entities in China included Cisco, Dell, HP, IBM, Intel, Yahoo and Manpower. For example, someone whose registration data suggests he’s in China registered kscisco.com, a doppelganger for ks.cisco.com. Another user who appeared to be in China registered nayahoo.com a variant of the legitimate na.yahoo.com (a subdomain for Yahoo in Namibia).
Kim said that out of the 30 doppelganger domains they set up, only one company noticed when they registered the domain and came after them threatening a lawsuit unless they released ownership of it, which they did.
He also said that out of the 120,000 e-mails that people had mistakenly sent to their doppelganger domains, only two senders indicated they were aware of the mistake. One of the senders sent a follow-up e-mail with a question mark in it, perhaps to see if it would bounce back. The other user sent out an e-mail query to the same address with a question asking where the e-mail had landed.
Defenses are few:
Companies can mitigate the issue by buying up any doppelganger domains that are still available for their company. But in the case of domains that may already have been purchased by outsiders, Kim recommends that companies configure their networks to block DNS and internal e-mails sent by employees that might get incorrectly addressed to the doppelganger domains. This won’t prevent someone from intercepting e-mail that outsiders send to the doppelganger domains, but at least it will cut down on the amount of e-mail the intruders might grab.
I suppose you can buy up the most common typos, but there will always be ones you didn’t think about—especially if you use a lot of subdomains.
There’s been a forged Google certificate out in the wild for the past month and a half. Whoever has it—evidence points to the Iranian government—can, if they’re in the right place, launch man-in-the-middle attacks against Gmail users and read their mail. This isn’t Google’s mistake; the certificate was issued by a Dutch CA that has nothing to do with Google.
This attack illustrates one of the many security problems with SSL: there are too many single points of trust.
EDITED TO ADD (9/1): It seems that 200 forged certificates were generated, not just for Google.
Sidebar photo of Bruce Schneier by Joe MacInnis.