New Draft Article: "Decryption Originalism: The Lessons of Burr"

What John Marshall, Aaron Burr, and some of the best lawyers of the founding era can teach us about the Fifth Amendment and compelled decryption.

Today I posted a new draft article on SSRN, Decryption Originalism: The Lessons of Burr. Here's the abstract:

The Supreme Court is likely to rule soon on how the Fifth Amendment privilege against self-incrimination applies to compelled decryption of a digital device. When the Court rules, the original understanding of the Fifth Amendment may control the outcome. This Article details an extraordinary case that illuminates the original understanding of the privilege and its application to compelled decryption. During the 1807 treason trial of Aaron Burr, with Chief Justice John Marshall presiding, the government asked Burr's private secretary if he knew the cipher to an encrypted letter Burr had sent to a co-conspirator. Burr's secretary pled the Fifth, leading to an extensive debate on the meaning of the privilege and an opinion from the Chief Justice.

The Burr dispute presents a remarkable opportunity to unearth the original understanding of the Fifth Amendment and its application to surprisingly modern facts. The lawyers in Burr were celebrated and experienced advocates. The Chief Justice allowed them to argue the Fifth Amendment question in exhaustive detail. And an attorney recorded the entire argument in shorthand, including dozens of legal citations to the specific pages of the authorities the lawyers invoked. The rich materials allow us to reconstruct for the first time precisely how the privilege was understood by leading lawyers and Chief Justice John Marshall soon after the Fifth Amendment's ratification. The Article presents that reconstruction, and it concludes by applying Burr's lessons to the modern problem of compelled decryption of digital devices such as cell phones and computers.

This is a new draft, so comments and criticism are particularly welcome. You can send them to me at orin [at] berkeley.edu.  Thanks for reading.

Comments

[apedad]

I guess the main difference I see in the Burr case and today's cases, is that Willie (who was pleading the 5th), was not Burr and yet in today's cases the defendant is in fact the person pleading the 5th.

Burr's case was decided because of the argument that Willie only needed to answer the question about encryption at the time of the trial (not during the alleged time of the criminal activity).

OK, that's hard to swallow but so be it.

I can't see how that could apply to today's cases.

We cannot compel a defendant - as a single individual - to divide their brain/memory into what they knew during the alleged criminal activity to what they know today; it's a continuum!

So to me, that's the main difference.

Also, you need to make this into a movie.

Treason, presidents, kings, plots, religion, law - a guaranteed hit!

[arpiniant1]

I'm not convinced
A cipher could have been known by many people and implies no knowledge or control, while a password almost always implies knowledge or control.

IOW, Burr could have written many documents in the same cipher, without his secretary ever seeing them. so knowledge of the cipher has no relation to knowledge of a crime.
Password implies knowledge and control of what is on the device, I would think a very different thing, unless it was a computer that many people had access and control of.

Additionally, what if the password was evidence in and of itself?
&mykiddiepornstash
or
*Mymurdervictimsname
or
#Treasonnotes

the authorities would not know that the password was evidence until they saw it. I do not believe entering the password oneself gets prosecutors off the hook, some devices will allow users to view the password once logged in

Face it, its an attempted cheat around the 5th amendment.

[Krayt]

There was considerable dispute in the Burr proceedings about how the prosecution obtained the letter. Burr’s counsel argued that the government stole it illegally from the postal mail.84 According to counsel, the theft was an unreasonable seizure that violated Burr’s Fourth Amendment rights and justified the letter’s suppression.85 The government denied the accusation.
86
The prosecution eventually produced a letter from the government’s chief witness, General James Wilkinson, claiming that a lawyer in New Orleans had given the letter to him

Defense: "If the lawyer had the letter, why didn't he have the cypher? Who was this lawyer? How did he work with you? How did he get it? Why would he even have it in his usual capacity as a lawyer?"

"We can't reveal the name of the whistleblower."

[A_Lior]

Page 54: "lesson if history" should be "lesson of history"

[Daniel R. Grayson]

This text appears more than once: "Because forcing Willie to testify about his present knowledge of the cipher would not prove his past knowledge of it, the answer could not incriminate Willie and it was Marshall’s domain to rule that Willie must answer."

Repeated "that" in "Marshall refers to “[a]uthorities” that that law lawyers “adduced”"

Repeated "that" in " and that that the prosecution’s failure to ask the court".

[anorlunda]

Does it make a different what crime may be exposed by giving the PW?

We suspect a person of having child porn on his computer. We force him to give the PW. No porn is found, but an admission of an unrelated crime (murder) is found. Can that evidence be used?

Kiddie porn fan hires a consultant to encrypt his data and pays a fee. Only the consultant has the PW. But consultant defrauded the porn fan. He never did archive the data, but deleted it instead. Consultant's fraud would be revealed if the PW is given up. What then?

[arpiniant1]

i think the pedo is unlikely to sue or press charges in this case....

On the other crimes front, I think once a crime is discovered, it takes little to connect the dots.

[DavidWH]

For people who are aware and motivated, a decision that passwords are not protected by the 5th amendment will be irrelevant.

For the sake of argument, assume that a password known by an individual is never protected by the 5th amendment and the court can order it to be revealed. The problem with this is that there are ways to save a password such that the individual never knows it, and revealing how to recover it plausibly destroys it. This can be done with things like code books, dominoes, playing cards, and serialized paper money. The use of this type of password system is inconvenient now but if pressed, it could be made much easier which would make existing password recovery for any purpose such as it is more difficult for the authorities.

[Peter Gerdes]

No, there is no system like you suggest.

Either the witness retains the ability to access the encrypted content or they do not. If they do not then there is no issue. If there is something the witness could do to recover the content then the court can just order the witness to go do it.

Now you can design a system with panic passwords or other ways to surreptitiously delete information. However, if you do that the court can just find you in contempt.

—

The best you can do technologically is to create a system in which the government can't prove there is encrypted content in the first place. For instance, you could fill half your drive with a combination of random cruft and encrypted content and have multiple passwords some of which decrypt more parts of the disk (but never all). Then the goverment wouldn't know if, in fact, there really was more information a password could be demanded for.

But that's all theory. In practice you can't make such a system userfriendly because it has the unfortunate feature that the system can't warn you if you're about to overwrite encrypted content rather than random noise and it is super hard to ensure no metadata proving the existence of further layers exists. Not to mention wasting tons of storage.

[DavidWH]

Here is a simplified example using bills:

I take a set of bills, say two $1s, two $5s, two $10s, two $20s, two $50s, and two $100s, shuffle them, and then use the last two characters of each serial number as the password so 12 digits. The search turns up this envelope of bills. Will they be kept in order after being seized? Will the $100s even be there when they deliver the envelope of bills to me so I can unlock the phone?

[Peter Gerdes]

Super interesting article but it seems to me there is a huge difference between asking Burr's secretary if he now knew the cipher and asking someone if they now know the password to an iphone just in terms of the plausibility of having only discovered the password after the illegal activity had taken place on the phone.

Burr's secretary really could have advanced a plausible argument, if tried, that Burr only told him the cipher much later so he could manage other correspondence. In contrast, were on I a jury and was told that the defendant testified that he knew the complex password to an iphone found in his house I'd be willing to credit the phone as the defendant's absent contrary evidence. Not to mention the fact that (should the actual password be disclosed) the fact that the defendant created it will generally be easy to determine by comparing it to passwords the defendant uses elsewhere, looking at word choices etcc.. even for complex passwords (if one person who might have created the file uses passphrases the other uses line noise passwords looking at the password tells us who created it even if both know the password).

[Peter Gerdes]

On a more general point I feel that this analysis of the lessons of the Burr trial is overly formalistic and thereby misses the bigger point.

If I take literally your interpretation of Marshal's rule I could create multiple questions, each of which easily avoids being barred by the 5th on your interpretation, yet together logically entail the defendant committed the crime.

For instance, both the question "Did you Willie learn the cipher after the date this letter was drafted" and "Do you Willie now know the cipher" both seem to pass the test you suggest. After all, no answer Willie could give to the first question could form a link in the sense you suggest since testifying that he didn't learn the cipher later doesn't rule out the possibility that he doesn't know the cipher *at all*. While the second question isn't problematic because Willie could have learned the cipher after the fact in an innocent fashion. Yet, if Willie had known the cipher when he copied the letter his testimony would logically entail he knew what the letter said when it was written…a question Marshal ruled he couldn't be forced to answer as it might incriminate him of the crime misprision of treason.

At a minimum this has some interesting consequences for conflicts between jurisdictions (what if one jurisdiction needs one answer and another the other but together they logically implicate the witness's 5th amendment right)? More broadly, I think it suggests that one can't really capture the intended protection of the 5th in such a deeply formal fashion.

I react rather differently to the historical material you covered. Perhaps this is incorrect but I was struck by the sense that everyone seemed to conceive of the 5th amendment as aimed toward ensuring people weren't put to a kind of Sophie's choice between damning themselves with admissions of guilt, perjury or contempt. I suggest this, more than a more formalistic analysis, is the more illuminating point. Moreover, it offers a better account that the choice between the purely causal analysis and some more specific 'link' to the crime.

After all, there is only a difference in degree between the way in which Willie's admission to knowing the cipher now suggests he may have known it when written (the difference being only how likely you think he learned it later is) and the way in which admitting knowledge of a password to contraband suggests past control over that contraband.

So no matter what you have to draw distinctions based on differences in degree and once you grant this isn't the simpler rule that the 5th amendment right attaches when the testimony is sufficiently likely (or feels that way) to substantially assist conviction that the defendant would feel put to an impossible choice. And this approach would say that not disclosing a password is near the core of the right the 5th protects when disclosing that password would essentially ensure conviction.

[Peter Gerdes]

The point about the multiple questions taken together is merely to suggest that the determination must involve a more holistic examination of likely litigation consequences and that no mere formal analysis of the relation between that admission and the elements of the crime can suffice (if Marshal's actual ruling was correct). While I'm sure you'll disagree I feel once you get to that point it makes it more appealing to instead evaluate the extent to which the question would put the witness in a (knowing?) bind between perjury and sending themselves to prison.

[FreedomTechie]

I'm not sure I recognize the distinction between knowledge of the cipher (in Burr's case) and knowledge of the key (or password used to derive it in modern cryptosystems). The notion of a secret key separate from a public algorithm was formalized in western academia in 1883 as Kerckhoffs's principle. But people had been separating keys from algorithms for far longer, with the best known example at the time being Bellaso's adaptation of Vignere's cipher, published in the 1550s and used by many militaries since.

Looking at the cipher used by Burr (https://books.google.com/books?id=T7NcAAAAcAAJ&lpg=PA215&dq=burr%20key%20cipher&pg=PA214#v=onepage&q=cipher&f=false), it used unique hieroglyphics as its ciphertext alphabet, which made it immediately identifiable as Burr's cipher, simply by noticing that a bunch of Burr's letters had the same hieroglyphics on them. But even in Burr's system, there's a clear specific "key" element to the algorithm - Burr used different words (like "CUBA" and "FRANCE") as different keys with his general-purpose cipher. The chosen word acted as a secret, flexible input from which the algorithm derived its security - exactly like modern password-based key derivation functions. Of course, Burr's cipher doesn't stand up to modern or even contemporaneous cryptanalytic techniques, but anyone except a cryptography expert could not decode one of Burr's messages with knowledge of the algorithm, but not his chosen password.

And modern cryptosystems pose far more features that stretch the analogy.

1. Many modern cryptosystems offer a property called Perfect Forward Secrecy - that property means that a specific communication cannot be decrypted after the fact, even if the keys of both parties to the communication are later compromised. As there is no way for a third party to become aware of the key/password later, demonstrating knowledge of or divulging the key is tantamount to admitting to being a party in the communication.

2. Multipurpose keys. Suppose a defendant intentionally creates a proof of some crime (for example, by signing a public defamatory statement or printable gun schematic with the same cryptographic key used to encrypt the file the government is interested in). Now compelling unlocking the file is equivalent to claiming authorship of the defamatory statement. This is somewhat equivalent to the "iloveburr" password mentioned in your paper, but is not mitigated by compelled entry rather than compelled disclosure.

3. Some software can be configured to provide different contents based on the password used to unlock it. In its default configuration, it behaves as if there are secondary passwords and contents, to provide deniability to users who do configure additional passwords. In this system, divulging the actual password or unlocking it is equivalent to disclosing the existence of a previously unknown repository of evidence or contraband.

[RandomWalk]

A couple of typos:

Page 33, last paragraph -- "asking a witness his confession was", should "confession" be "profession" ?

Page 48, last paragraph -- "unlawful to important, export, or possess", "important" should be "import".