Saturday, November 16, 2013

Our Government Has Weaponized the Internet. Here’s How They Did It





The internet backbone — the infrastructure of networks upon which internet traffic travels — went from being a passive infrastructure for communication to an active weapon for attacks.
According to revelations about the QUANTUM program, the NSA can “shoot” (their words) an exploit at any target it desires as his or her traffic passes across the backbone. It appears that the NSA and GCHQ were the first to turn the internet backbone into a weapon; absent Snowdens of their own, other countries may do the same and then say, “It wasn’t us. And even if it was, you started it.”
If the NSA can hack Petrobras, the Russians can justify attacking Exxon/Mobil. If GCHQ can hack Belgicom to enable covert wiretaps, France can do the same to AT&T. If the Canadians target the Brazilian Ministry of Mines and Energy, the Chinese can target the U.S. Department of the Interior. We now live in a world where, if we are lucky, our attackers may be every country our traffic passes through except our own.
Which means the rest of us — and especially any company or individual whose operations are economically or politically significant — are now targets. All cleartext traffic is not just information being sent from sender to receiver, but is a possible attack vector.
Here’s how it works.
The QUANTUM codename is deliciously apt for a technique known as “packet injection,” which spoofs or forges packets to intercept them. The NSA’s wiretaps don’t even need to be silent; they just need to send a message that arrives at the target first. It works by examining requests and injecting a forged reply that appears to come from the real recipient so the victim acts on it.
In this case, packet injection is used for “man-on-the-side” attacks — which are more failure-tolerant than man-in-the-middle attacks because they allow one to observe and add (but not also subtract, as the man-in-the-middle attacks do). That’s why these are particularly popular in censorship systems. It can’t keep up? That’s okay. Better to miss a few than to not work at all.
The technology itself is actually pretty basic. And the same techniques that work on on a Wi-Fi network can work on a backbone wiretap. I personally coded up a packet-injector from scratch in a matter of hours five years ago, and it’s long been a staple of DefCon pranks.
So how have nations used packet injection, and what else can they do with it? These are some of the known uses.

Censorship

The most infamous use of packet injection prior to the Snowden leaks was censorship, where both internet service providers (ISPs) and the Great Firewall of China injected TCP reset packets (RST) to block undesired traffic. When a computer receives one of these injected RST packets, it closes the connection, believing that all communication is complete.
Although public disclosure forced ISPs to stop this behavior, China continues to censor with injected resets. It also injects the Domain Name System (DNS) — the system all computers use to turn names such as “www.facebook.com” into IP addresses — by inserting a fake reply whenever it sees a forbidden name. (It’s a process that has caused collateral damage by censoring non-Chinese internet traffic).

User Identification

User cookies, those inserted by both advertising networks and services, also serve as great identifiers for NSA targeting. Yet a web browser only reveals these cookies when communicating with such sites. A solution lies in the NSA’s QUANTUMCOOKIE attack, which they’ve utilized to de-anonymize Tor users.
A packet injector can reveal these cookies by replying to an unnoticed web fetch (such as a small image) with a HTTP 302 redirect pointing to the target site (such as Hotmail). The browser now thinks “hey, should really go visit Hotmail and ask it for this image”. In connecting to Hotmail, it reveals all non-secure cookies to the wiretap. This both identifies the user to the wiretap, and also allows the wiretap to use these cookies.
So for any webmail service that doesn’t require HTTPS encryption, QUANTUMCOOKIE also allows the wiretap to log in as the target and read the target’s mail. QUANTUMCOOKIE could also tag users, as the same redirection that extracts a cookie could also set or modify a cookie, enabling the NSA to actively track users of interest as they move across the network — although there is no indication yet that the NSA utilizes this technique.

User Attack

The NSA has a collection of FOXACID servers, designed to exploit visitors. Conceptually similar to Metasploit’s WebServer browser autopwn mode, these FOXACID servers probe any visiting browser for weaknesses to exploit.
All it takes is a single request from a victim passing a wiretap for exploitation to occur. Once the QUANTUM wiretap identifies the victim, it simply packet injects a 302 redirect to a FOXACID server. Now the victim’s browser starts talking to the FOXACID server, which quickly takes over the victim’s computer. The NSA calls this QUANTUMINSERT.
The NSA and GCHQ used this technique not only to target Tor users who read Inspire (reported to be an Al-Qaeda propaganda magazine in the English language) but also to gain a foothold within the Belgium telecommunication firm Belgicom, as a prelude to wiretapping Belgium phones.
One particular trick involved identifying the LinkedIn or Slashdot account of an intended target. Then when the QUANTUM system observed individuals visiting LinkedIn or Slashdot, it would examine the HTML returned to identify the user before shooting an exploit at the victim. Any page that identifies the users over HTTP would work equally well, as long as the NSA is willing to write a parser to extract user information from the contents of the page.
Other possible QUANTUM use cases include the following. These are speculative, as we have no evidence that the NSA, GCHQ, or others are utilizing these opportunities. Yet to security experts they are obvious extensions of the logic above.
HTTP cache poisoning. Web browsers often cache critical scripts, such as the ubiquitous Google Analytics script ‘ga.js’. The packet injector can see a request for one of these scripts and instead respond with a malicious version, which will now run on numerous web pages. Since such scripts rarely change, the victim will continue to use the attacker’s script until either the server changes the original script or the browser clears its cache.
Zero-Exploit Exploitation. The FinFly “remote monitoring” hacking tool sold to governments includes exploit-free exploitation, where it modifies software downloads and updates to contain a copy of the FinFisher Spyware. Although Gamma International’s tool operates as a full man-in-the-middle, packet injection can reproduce the effect. The injector simply waits for the victim to attempt a file download, and replies with a 302 redirect to a new server. This new server fetches the original file, modifies it, and passes it on to the victim. When the victim runs the executable, they are now exploited — without the need for any actual exploits.
Mobile Phone Applications. Numerous Android and iOS applications fetch data through simple HTTP. In particular, the “Vulna” Android advertisement library was an easy target,  simply waiting for a request from the library and responding with an attack that can effectively completely control the victim’s phone. Although Google removed applications using this particular library, other advertisement libraries and applications can present similar vulnerabilities.
DNS-Derived Man-in-the-Middle. Some attacks, such as intercepting HTTPS traffic with a forged certificate, require a full man in the middle rather than a simple eavesdropper. Since every communication starts with a DNS request, and it is only a rare DNS resolver that cryptographically validates the reply with DNSSEC, a packet injector can simply see the DNS request and inject its own reply. This represents a capability upgrade, turning a man-on-the-side into a man-in-the-middle.
One possible use is to intercept HTTPS connections if the attacker has a certificate that the victim will accept, by simply redirecting the victim to the attacker’s server. Now the attacker’s server can complete the HTTPS connection. Another potential use involves intercepting and modifying email. The attacker simply packet-injects replies for the MX (Mailserver) entries corresponding to the target’s email. Now the target’s email will first pass through the attacker’s email server. This server could do more than just read the target’s incoming mail, it could also modify it to contain exploits.
Amplifying Reach. Large countries don’t need to worry about seeing an individual victim: odds are that a victim’s traffic will pass one wiretap in a short period of time. But smaller countries that wish to utilize the QUANTUMINSERT technique need to force victims traffic past their wiretaps. It’s simply a matter of buying the traffic: Simply ensure that local companies (such as the national airline) both advertise heavily and utilize in-country servers for hosting their ads. Then when a desired target views the advertisement, use packet injection to redirect them to the exploit server; just observe which IP a potential victim arrived from before deciding whether to attack. It’s like a watering hole attack where the attacker doesn’t need to corrupt the watering hole.
***
The only self defense from all of the above is universal encryption. Universal encryption is difficult and expensive, but unfortunately necessary.
Encryption doesn’t just keep our traffic safe from eavesdroppers, it protects us from attack. DNSSEC validation protects DNS from tampering, while SSL armors both email and web traffic.
There are many engineering and logistic difficulties involved in encrypting all traffic on the internet, but its one we must overcome if we are to defend ourselves from the entities that have weaponized the backbone.

No comments: