Recently, on a 16-person group chat initiated by Michael Waltz and including Pete Hegseth, Jeffrey Goldberg — editor-in-chief of The Atlantic and a vehement critic of both Trump and Hegseth — was inconveniently included. Goldberg stayed on the chat and took screenshots, then withheld reporting the story until, conveniently, the night before Waltz and Hegseth were to testify before Congress.
Furor followed, with Goldberg insisting that highly secret information was disclosed in the chat, while primly insisting that he wouldn’t disclose what it was — until challenged, at which point it appeared that his skills as a classification officer were lacking.
Democrats used the furor to attack both Hegseth and Waltz, largely at the hands of people who had been critical of Hegseth’s nomination. However, Trump and Waltz insist that the information, while sensitive, wasn’t classified, and targeting Hegseth — who didn’t initiate the chat — seems to primarily be political.
Still, it is embarrassing, and the accidental inclusion of Goldberg seems… odd.
As more information has come out, the explanation has been that Waltz had Goldberg’s phone in his contacts under the name of Brian Hughes, the current National Security Council spokesperson, apparently thanks to his iPhone helpfully adding the number via communications between Waltz and Hughes.
Still, it’s a security violation, albeit one that in the old days we would have called a “minor violation” (I understand that’s outmoded and there’s a more complicated system now, the details of which are apparently classified.).
There’s also the fact that the government, through CISA, had recommended Signal for private communications.
This is the crux of the problem. What happened was not a failure of Signal — it did what it’s advertised to do: provide end-to-end encrypted communication. The failure was, rather, that unsophisticated users were misled that privacy was the same as security.
The missing step there is that while the communications are private, Signal only authenticates the user endpoint that they have the phone number that was added to the chat. In this case, Goldberg was added to the chat because his phone was in Waltz’s phone book. Signal says, “You want to talk to 867-5309? You bet.”
Failing to make sure the connection was authentic to Hughes and not some random Trump administration critic is where it fell down.
I imagine that everyone involved is appropriately embarrassed, and I’m sure many forceful memos are going out. I wouldn’t be too surprised if CISA’s recommendation of Signal has added to it something strongly reminding users that just because it’s a private communication doesn’t mean that it’s a secure channel.
That’s just a band-aid, however. Using really secure communications is what we in the computer security world call “a pain.” (The actual technical term is more definite, but you get the idea.) Pushing people to use SIPRnet, or worse, JWICS, would mean communications could only be done from a secured facility — a SCIF — or its equivalent. (You can bet that Air Force One is treated as a SCIF.)
The thing is, if you make security inconvenient enough, people will find ways to circumvent it. It would be better if the government provided an application for private and authenticated communication.
It turns out that Signal, since it’s open source, would be a good start. Call it “Signal: Fort Meade Edition.” It would be a fork of the existing Signal, ideally also open-source, with a couple of extensions.
First, add a layer of identification to the endpoint devices. This is already part of, for example, the TLS protocol, which is what you’re using whenever you have a URL that starts with “https:.” It depends on providing a “cert,” an X.509v3 cryptographic certificate that uses a cryptographic signature to provide authenticity. On the web, this is commonly only at the server end, but the protocol provides for certificates at both ends. This would provide some confidence that the endpoint devices were authorized devices. Not perfect, but a step in the right direction.
Second, provide users with a PIN, either a six-digit number or a pronounceable nonsense word. The nonsense word is probably better; it’s easier to communicate. But what’s essential is that the PIN is communicated “out of band.” You don’t send it with the Signal channel, and you either change it daily or even better, you change it for each chat, communicating the intention to have a chat at a certain time using a certain PIN using a more secure channel like a SIPRnet email.
In use, it would be very much like Signal is now. Signal already establishes a session key for the encryption using Double-Ratchet encryption, an extension of the well-known Diffie-Hellman algorithm. At the time of establishing the chat, each user would have to provide the day’s PIN before they could make the connection.
The underlying issue, though, is that people, even users who work with classified data and the normal classified channels, aren’t very sophisticated in the concepts involved. This kerfuffle was caused because the people using Signal accepted CISA’s assurance and recommendation that Signal was good for private communications and were fooled into thinking it was enough for secure communications.
Editor’s Note: Do you enjoy PJ Media’s conservative reporting taking on the radical left and woke media? Support our work so that we can continue to bring you the truth. Join PJ Media VIP and use the promo code FIGHT to get 60% off your VIP membership!
