#TeleMessage customers include DC Police, Andreessen Horowitz, JP Morgan, and hundreds more
#signal

I've been digging through the 410 GB of Java heap dumps from TeleMessage's archive server, provided by DDoSecrets. Here's a description of the dataset, some of my initial findings, details about an upcoming open source research tool I'm going to release, and a huge list of potential TeleMessage customers.

I found a WhatsApp group called "MPD Command Staff" with 46 users in it. There are many messages in this group, but they're all encrypted. (As I described in my earlier analysis, some of the individual messages are encrypted.) I looked up some of the phone numbers from this group on OSINT Industries and quickly discovered that these people all work for the Metropolitan Police Department in Washington, DC.

I also found a message sent to a Signal group called "US / China AI Race." The Signal group had 100 people in it. I looked some of them up: many of the group members hold prominent positions at major universities, the defense industry, and the military, and all seem to do AI-related work. The message says, "The biggest crime was USG ignored these fabs for two years." That's it. The dataset doesn't include any other messages from this Signal group.

That said, here's what I've found:
60,012 messages.
36,388 of the messages are plaintext, and 23,624 are encrypted.
1,079 of the message include full attachments (like images, videos, PDFs, contact files, etc.) that are actually part of the dataset. But of those, only 50 of messages are in plaintext. I can, however, actually open and view those plaintext attachments.
Most messages have a subject field that's something like, "WhatsApp message from X to Y." Based on these subjects:
37,753 are WhatsApp messages.
2,549 are Telegram messages.
455 are SMS messages.
141 are Signal messages.
95 are something called "App Messages."
26 are MMS messages.
26 are WeChat messages.
16 are voice calls logs.
11,254 are missing subject fields.
3,501 group chats, the vast majority of which are WhatsApp.
At least 2,034 are WhatsApp groups.
At least 578 are SMS groups.
At least 256 are Telegram groups.
At least 26 are Signal groups.
At least 10 are WeChat groups.
I'm not sure about the other ~600 groups, though it's possible to determine by manually looking at the messages associated with them.
There are also plenty of individual messages that are clearly part of a group chat, but that didn't include JSON metadata related to it, so they're not categorized as groups, even though they are.
44,503 users. These are either senders or recipients of messages.
At least 25,792 of them use phone numbers as the identifier.
At least 31 of them use email addresses, and at least 391 look like they use usernames.
I'm not sure about another 18,289 of them, but I think most of them are also phone numbers.
17,377 of them include first and/or last names, too.

https://micahflee.com/telemessage-customers-include-dc-police-andreesen-horowitz-jp-morgan-and-hundreds-more/

https://micahflee.com/telemessage-explorer-a-new-open-source-research-tool/