Electronic surveillance succeeds because it is secret. When the government seeks to record “what is whispered in the closet,” in the words of Justice Brandeis, it must use clandestine methods. Since at least 1928, when Brandeis wrote his United States v. Olmstead dissent, it has been understood that unseen surveillance also provides a “subtler and more far-reaching means of invasion of privacy” than physical searches. Recognition of this dual nature — effective but invasive — has driven evolution of the law. From Berger v. New York, United States v. Katz, and the Keith case, to recent decisions like United States v. Jones, the Supreme Court has emphasized the need to control electronic surveillance, cabining executive discretion and requiring a highly particularized showing of what will be searched or seized before a warrant is issued. Congress has in turn responded with laws like Title III, ECPA, and FISA.
Even so, Brandeis made an easy bet when he predicted that “[t]he progress of science in furnishing the Government with means of espionage is not likely to stop with wire-tapping.” As Ahmed Ghappour explained in 2014, the state of the government’s art is now hacking: “remote access of a computer to install malicious software without the knowledge or permission of the owner/operator.” In the computer security field, this sort of malware is often called a RAT, for Remote Access Tool, but the FBI and DOJ refer to it as a NIT, for Network Investigative Technique. The FBI has used NITs and similar malware for more than 15 years, but it is only now that a significant number of criminal cases involving government hacking techniques are being litigated. Just as with wiretapping, we should be mindful of the need for both constitutional and statutory law to keep up with the use of hacking for surveillance.
The bulk of the recent NIT cases arise from a single government investigation, in which the FBI took over the server hosting a Tor hidden service website called Playpen that featured child pornography. The properties of Tor often make it difficult to identify users of a hidden service site, but by seizing the website’s server, the FBI was able to take advantage of a vulnerability in software running on users’ computers to covertly install a NIT to obtain their IP addresses and other identifying information. In February 2015, a federal magistrate judge in the Eastern District of Virginia granted the FBI a single warrant to run a watering-hole attack and install NITs on computers belonging to “any user or administrator” of the site. In the two weeks the FBI ran the site, NITs identified thousands of users, more than 130 of whom have been prosecuted so far.
Much of the attention paid to these cases has focused on whether the warrant issued by the magistrate judge in Virginia satisfies Rule 41(b) of the Federal Rules of Criminal Procedure, which places clear restrictions on magistrate judges’ authority to issue warrants to search property located outside of their districts. Contrary to Judith Germano’s claim that the warrant was proper under Rule 41, every court I am aware of that has considered the issue has determined that the rule was violated. (Some have determined that the violation should not give rise to a suppression of evidence, however.)
Not coincidentally, in late 2014, the Justice Department proposed — and the Judicial Conference and Supreme Court have approved — an amendment to Rule 41 that would allow magistrate judges to issue warrants for remote hacking of unknown computers in any district, if users have concealed their locations “through technological means.” Unless Congress affirmatively acts, the changes will go into effect on December 1. Following opposition by civil liberties groups, however — including my employer, the Electronic Frontier Foundation — Senator Ron Wyden has introduced the cleverly acronymed SMH (Stopping Mass Hacking) Act of 2016 to stop the amendments, and a bi-partisan group of Representatives have introduced similar legislation in the House.
The Fourth Amendment and NITs
Even if Rule 41 is amended to allow the issuance of NIT warrants, courts must contend with the considerable Fourth Amendment concerns raised by remotely hacking unknown numbers of users.
As a preliminary matter, each use of a NIT appears to involve at least three Fourth Amendment “events” that implicate users’ constitutional interests:
- The installation of a NIT is a seizure because the government interferes with the user’s “possessory interest” in their computer and exercises “dominion and control” over it under the applicable test from United States v. Jacobsen.
- Operation of the NIT to determine the computer’s assigned IP address and other information is a search, since it requires examining information over which the user has a “reasonable expectation of privacy.”
- Finally, the copying of this information is a second seizure, since it interferes with the user’s “dominion and control” over that information.
Nevertheless, some courts discussing the Fourth Amendment implications of NIT warrants have looked to the information obtained by the NIT without also considering how and where the government got this information. For example, in United States v. Werdene, the court applied Third Circuit precedent in holding that users do not have a reasonable expectation of privacy in their IP addresses because they share them with third parties. But the government used NITs in the Playpen investigation precisely because Tor obscured users’ IP addresses from the server controlled by the government. By employing the NIT on users’ computers, the government forced users to disclose their IP addresses (along with other private information). It is therefore irrelevant that IP addresses are conveyed to third parties in other contexts, since the government activity at issue occurred in a space where defendants have a reasonable expectation of privacy — their computers.
The Particularity Problem with NITs
Defenders of the government’s techniques like Germano are quick to point out that, Rule 41 aside, the government apparently routinely obtains warrants to install NITs. (It likewise reportedly applies for Title III warrants in cases where it goes beyond the collection of IP addresses and uses them for active eavesdropping.) Germano and others argue this is sufficient for Fourth Amendment purposes.
But there is good reason to be skeptical that the NIT warrants we’ve seen can satisfy the Fourth Amendment. The Warrant Clause requires not only that warrants be supported by probable cause, but that they “particularly describ[e] the place to be searched, and the persons or things to be seized.” By contrast, the Playpen NIT warrant authorized the search and seizure of information located on unknown computers in unknown places belonging to any and all users of the site (and there were over 150,000 of them in this case).
The Fourth Amendment requires more. For example, as the Supreme Court explained in Steagald v. United States, arrest warrants must specify particular houses if police seek to enter to make the arrest, because they cannot leave “to the unfettered discretion of the police the decision as to which particular homes should be searched.” Warrants that lack sufficient particularity instead give rise to “all-encompassing dragnets.” The Playpen NIT warrant vested near total discretion in the FBI as to which users it would search and by what means.
Defenders of NITs argue that when it comes to identifying individuals who have taken steps to conceal themselves, such as by using Tor, the government is simply unable to include any more specificity in a warrant and that the Fourth Amendment is flexible enough to allow such vagueness. In the Playpen cases, this is incorrect on both factual and legal grounds.
First, by operating the site for two weeks, the FBI had extensive information about users’ activity and could have proceeded in a less sweeping fashion by obtaining narrower warrants that named specific users and offered specific facts describing their connection to the site.
Second, while some courts have at times allowed “roving wiretaps” (which name specific persons but not places) and “all persons” warrants (which name specific places but not specific persons), no court has previously upheld the issuance of a warrant to search unknown persons in unknown places. Without particularity on either front, this NIT warrant was akin to a dragnet for all users of the site.
Title III for Government Hacking
The particularity problem encountered in the Playpen cases is only the latest evidence of the inherent tension with surveillance identified by Justice Brandeis in Olmstead. In Berger, the Supreme Court explained that because eavesdropping involves “an invasion of privacy that is broad in scope,” the Fourth Amendment requires “precise and discriminate procedures” for issuing warrants. For wiretapping, this means that conversations to be recorded must “be described with particularity.” Passed in response to Berger, Title III of the Omnibus Crime Control and Safe Streets Act of 1968 sets forth these procedures, requiring extensive particularity showings and a demonstration that other methods of investigation will be ineffective, among other safeguards.
The remote installation of malware needs similar protections. Remote hacking is at least as invasive as wiretapping, even when the information ultimately seized is relatively limited. The installation of malware, even by a government actor, involves a number of significant and even potentially dangerous steps, including the use of unpatched vulnerabilities on a user’s computer and altering code running on that computer. These actions may leave users vulnerable to malicious third parties who also know of the vulnerability or who can hijack the NIT. And since the government refuses to reveal how it exploits vulnerabilities to install NITs, it is impossible for these vulnerabilities to be fixed or for defendants charged with evidence obtained from NITs to raise an effective defense. Last week, a district court in Washington recognized this dilemma and excluded the evidence against a Playpen user as a consequence of the government’s refusal to reveal its exploit.
Particularly in the absence of a statutory structure like Title III, courts are empowered to set limits on remote hacking warrants. For example, in the NIT context, a court could require that a warrant:
- tailor the search to identifying only the IP address (or related information) of a specifically targeted user;
- demonstrate that other avenues for identifying the physical location or actual identity of the user were undertaken and failed;
- cabin the discretion of the officers executing the warrant by establishing particular criteria under which the searches and seizures could occur; and
- require that targets of the NIT be notified afterward.
Applications that do not include these safeguards should be denied, as Magistrate Judge Smith of the Southern District of Texas did in 2013 when considering a constitutionally deficient application to install malware on a suspect’s computer.
* * *
These are far from the only problems with the government’s use of malware. For example, it sometimes uses previously unknown vulnerabilities or “zero days” in both law enforcement and intelligence operations. And even though the NSA and the White House agree that vulnerabilities should be disclosed in most cases, the government apparently prefers to let defendants go free rather than disclose the information under seal, just as it has done before with Stingrays. Thus, a comprehensive solution to government hacking should also impose protocols on the government for demonstrating that its need to exploit vulnerabilities for law enforcement or intelligence purposes in each instance outweighs the harm to the public in letting these vulnerabilities go unpatched.
Of course, the imposition of Title III-esque safeguards on remote hacking won’t solve all of these problems, but it will begin to address the greatest of the Fourth Amendment concerns.