Reading width
Wide uses the full column for everything, text, diagrams, code, and exercises. Narrow keeps the standard reading width.
Text size
Scales the body text. Headings and code blocks keep their size.
In this section
Browser and Email Forensics on Windows: Firefox Form History, Chrome Downloads and the Mailbox
Introduction
Most of what a person does on a computer happens inside a few applications: a browser or two, a mail client, a chat app. Each keeps its own databases in the user's profile, in its own format, and none of them is part of Windows. They hold what the operating system's artifacts cannot: what the person searched for, which pages they read, who they wrote to and what they attached.
In an insider case they often hold the evidence of intent.
Two browsers in one profile on one workstation. The everyday browser held the channel's setup; the second held the searches.
This lesson reads the insider's two browsers and mailbox, and places what they hold on the timeline lesson 0.6 built. Module 10 teaches each application family in depth: Chromium and Firefox, private browsing, Electron apps such as Teams and Slack, the Windows Search database, email on the endpoint, and cloud application artifacts.
Two Browsers, Two Records
Why a second browser mattersThe workstation had two browsers, and they held different things. Chrome was the everyday browser: 467 history entries dominated by company SharePoint, Outlook on the web, developer documentation and the usual mix of news and shopping. Firefox was used far less, 50 history entries, and its search bar held a short list of things the user typed.
Each browser keeps its data in a profile folder in the user's AppData. The Chromium project documents Chrome's default location on Windows as the User Data folder under the local app data folder, with each profile a subfolder inside it, often named Default.
Mozilla documents Firefox's own profile files: places.sqlite holds bookmarks, downloads and visited sites, and formhistory.sqlite holds what was typed into the search bar and into forms on websites.
Why the user had two browsers is not something the records answer. Many people keep a second browser for personal use, for testing, or out of habit. Here the second browser holds almost nothing but these searches and some developer pages, and the report can state that pattern without guessing at the motive behind it.
An investigation that collected one browser and assumed it was the only one would have missed the searches entirely. The first step with any user profile is to list every browser and every profile inside each, because people do separate activities by browser, sometimes deliberately. Module 10's Firefox lesson is largely about analysis across several browsers at once for this reason.
Before reading any browser, find where each keeps its data:
Where each browser keeps it
Collect all of themBoth browsers store their records as SQLite databases, which any SQLite reader opens. Dedicated browser parsers add the conversions that matter, such as each browser's own timestamp format into UTC, and they are what Module 10 uses for every browser it reads.
Firefox's profile sits under the roaming application data folder rather than the local one, which matters for collection: a triage target that gathers only local application data misses it. A collection for browsers takes both folders.
What Was Searched
Firefox form historyThe Firefox search bar's history, as the profile stored it:
First used (UTC) Times Value
-------------------- ----- -----------------------------------------
2026-01-20T10:11:22 3 css grid vs flexbox
2026-01-25T15:22:33 2 react useEffect cleanup
2026-02-03T11:55:11 1 python virtualenv windows
2026-02-12T14:15:22 1 bypass DLP monitoring corporate
2026-02-12T14:18:44 1 how to avoid network monitoring at work
2026-02-12T14:22:11 1 personal hotspot bypass corporate firewall
2026-02-12T14:25:33 1 does dropbox encrypt uploads
2026-02-13T09:33:11 2 mobile hotspot windows 10 connect
Eight entries, and they fall into two groups. Three are ordinary development questions spread across January and February, the kind of thing an engineer looks up while working. Five are not.
On 12 February, in about ten minutes, the user searched for how to bypass DLP monitoring, how to avoid network monitoring at work, how a personal hotspot could get around a corporate firewall, and whether Dropbox encrypts uploads. On 13 and 14 February, the same profile searched twice for how to connect a mobile hotspot to Windows.
The counts separate the two groups as clearly as the wording. Two of the development questions were used two or three times over several weeks, the pattern of someone returning to a reference while working. Each monitoring search was used exactly once, all in one sitting, and the hotspot question twice on consecutive days.
Ordinary work recurs; a specific question asked once and then acted on does not need asking again.
The form history table records each value with the number of times it was used and the first and last time. That is all it stores: not the results, not which link was clicked, not how long anyone spent reading. The Firefox history for those days would show which pages followed, and Module 10 reads the two together.
The survival of these entries is itself worth noting. Firefox lets a user clear form and search history along with browsing history, and the user here did not, or did not in this browser. People who prepare carefully in one respect often leave records in another, which is the same pattern lesson 0.8 found in the ransomware attacker.
Mozilla's description of the file is worth holding onto: it remembers what was searched for in the search bar. Values typed into a website's own search box are stored there too, as form entries, which is why a user who searches on a search engine's page rather than in the browser's bar can still leave a record.
Placing It on the Timeline
Between the setup and the first useTaken alone, a search for how to avoid monitoring could be curiosity. Plenty of people search for security topics for reasons that have nothing to do with their own plans, and an investigator who treats every such search as a confession will be wrong often. Its weight comes from where it sits on the timeline lesson 0.6 built:
The browser evidence on the case timeline. The searches sit between the channel being set up and its first use, three days before the first upload.
The channel was set up on 22 January: Dropbox installed and configured with a personal account, the USB drive connected for the first time. Three weeks later, the searches asked how to get around exactly the monitoring that would have caught an upload, and whether the upload itself would be encrypted.
The next two days asked how to connect a hotspot. On 15 February the hotspot profile was created, and Dropbox sent its first 1.25 GB over it the same afternoon.
Set out with the record each line comes from:
2026-01-22T09:16 Chrome Dropbox installer downloaded
2026-01-22T09:24 Dropbox client personal account configured
2026-02-12T14:15 Firefox first monitoring search
2026-02-12T14:25 Firefox does dropbox encrypt uploads
2026-02-13T09:33 Firefox mobile hotspot windows 10 connect
2026-02-15T14:18 Network list Personal-Hotspot-5G created
2026-02-15T14:20 SRUM Dropbox sends 1,247,000,000 bytes
Each step is recorded by a different program: Chrome, Firefox, the Dropbox client, the Windows network list and SRUM. None of them knew about the others. The sequence they make, setup, research, then use of exactly the method researched, is what the case brief's question about deliberate activity needs.
The three weeks between setup and research are quiet in the records. Dropbox synced small amounts over the corporate network in that period, as lesson 0.6 showed, and the USB drive was not connected again. That quiet period is itself part of the pattern: the tools were in place, and nothing was taken until after the searches.
The other end of the timeline is fixed by lesson 0.6: the last upload on 5 March, two days before the user's last day. The browser evidence therefore sits near the beginning of a six-week sequence whose end is also recorded, and the report can describe the whole of it, from setup to last use, with each step attributed to the record that holds it.
The hotspot searches are the most specific link in the whole chain. They ask how to do the thing that the network list records happening the next day, on a profile that never appears on any day without an upload.
What a Search Proves
Intent, sized to the evidenceSearch history is persuasive, which is exactly why it needs care in a report:
What a search proves
Intent, sized to the evidenceThe record proves that the words were typed into this Firefox profile's search bar at these times. It does not prove the user read anything or acted on it, and on its own it cannot say who was at the keyboard: that is the attribution question lesson 0.3 set out, and the same answer applies.
The profile belongs to the user's account, and the 12 February searches fall inside an interactive session that account opened at 08:03 that morning.
What gives the searches their weight here is not their wording but their fit. A report that says the user searched for how to bypass monitoring, three days before using a hotspot to upload to a personal account, and searched for how to connect that hotspot the day before first joining it, states facts that each record supports.
A report that says the user planned to evade detection draws the conclusion for the reader. The first is stronger in front of anyone who will challenge it, because each clause can be checked.
Wording matters at the level of single verbs. Searched for is what the record shows. Researched implies reading, which the record does not show. Planned implies intent formed in advance, which is an inference from the sequence. A report can use the stronger words where the evidence supports them, but each step up should be visible to the reader as a step, with the evidence that justifies it.
There is also an innocent reading to address. Engineers search for many things, and a single search for a security term proves nothing. Five searches in two days, on one topic, immediately before that topic's method was used, is the pattern that rules the innocent reading out, and the report says so rather than leaving the reader to infer it.
Two practical points follow for collection. Form history lives in its own file, separate from the history database, so a collection that took only places.sqlite would have missed it. And browsers hold these databases open while running, so they are best collected from an image or with a tool that copies locked files, as lesson 0.1's KAPE run did for the registry hives.
The Download That Set It Up
Chrome, beside the client it installedThe Dropbox installer's download record, read beside the client configuration it led to:
Started 2026-01-22T09:16:33Z
Saved to C:\Users\jmorrison\Downloads\DropboxInstaller.exe
From https://www.dropbox.com/download?plat=win&type=full
Referrer https://www.dropbox.com/install
Bytes 45,672,893
Dropbox client config created 2026-01-22T09:24:11Z, jmorrison.personal@gmail.com
The download began at 09:16:33 from Dropbox's own site, the referrer being its install page. Lesson 0.6 had the installer running from Prefetch at 09:17 and the client configured with the personal account at 09:24:11. The browser record adds the first step: the user went to Dropbox's site and downloaded the installer, rather than it arriving by some other route.
The last two entries in the list are documents rather than installers: meeting notes and a spreadsheet of bridge calculations, downloaded in February. For an engineer on a bridge project that is ordinary, and nothing else in the records connects either file to the upload days. They are noted and set aside, which is the treatment every unconnected record gets, rather than being stretched to fit the case.
The referrer field records the page the download was started from, which is how the record shows the user was on Dropbox's install page rather than following a link from an email or another site. A download whose referrer is a mail service or an unfamiliar domain tells a different story, and lesson 0.4's payload, with a lookalike domain as its referrer, was exactly that.
Chrome's history format is shared by every Chromium browser, including the Edge profile lesson 0.4 read on the ransomware host. The same database tables hold visits and downloads, with the same fields, which is why one parser and one method cover both browsers. Module 10's Chromium lesson reads the rest of the profile: sessions, the cache, autofill and the records that survive a history clear.
The full downloads list in Chrome is short:
2026-01-18T16:33:44 python-3.12.2-amd64.exe 26,543,210
2026-01-22T09:16:33 DropboxInstaller.exe 45,672,893
2026-01-28T14:22:33 VSCodeSetup-x64-1.96.exe 98,765,432
2026-02-05T09:11:22 zoom_installer.exe 34,567,890
2026-02-10T11:33:22 meeting_notes_2026-02-10.pdf 245,000
2026-02-18T15:44:11 bridge_calcs_v3.xlsx 1,456,000
Not every download is relevant, and the list shows how few are. Python, Visual Studio Code and Zoom installers and two work documents are ordinary for an engineer. The Dropbox installer is the one that connects to the case, and it is identified by its place on the timeline rather than by anything suspicious about the file itself.
The Mailbox
What it shows, and what it rules outEmail is the channel most people expect data to leave by, so it is checked even when another channel has already been found:
Reading a mailbox in an exfiltration case. The last row is a finding too, and often the one a reader asks about first.
The folder summary, read first, gives the scale of the mailbox and shows whether anything unusual was filed or deleted:
Inbox 3421 items 47 unread
Sent Items 1247 items 0 unread
Deleted Items 892 items 12 unread
Drafts 3 items 3 unread
Junk Email 234 items 234 unread
Archive 5621 items 0 unread
Projects/NGE-Bridge-Phase2 312 items 0 unread
Projects/NGE-TunnelDesign 187 items 0 unread
Projects/ClientReviews 445 items 2 unread
The user's sent mail in the final week, by recipient:
2026-03-02T11:55 a.kumar@northgateeng.com attachment
2026-03-03T09:11 engineering-team@northgateeng.com
2026-03-03T14:22 p.greaves@northgateeng.com
2026-03-04T11:33 h.wright@northgateeng.com attachment
2026-03-04T15:22 engineering-team@northgateeng.com attachment
2026-03-04T16:45 r.okafor@northgateeng.com
2026-03-05T09:22 p.chen@northgateeng.com
2026-03-05T10:14 engineering-team@northgateeng.com attachment
Every message went to a company address, all 8 of them. The attachments, a reinforcement schedule, a project update, handover calculations and a CAD file, went to colleagues and to the team, in messages whose subjects are about handover.
That is what a departing engineer's last week looks like, and it is evidence about the channel: email was available and the user did not use it to send data outside the company.
The largest attachment in the week was a CAD file of about 8.8 MB to a colleague, which is small beside the gigabytes that left by Dropbox. That comparison is worth making in the report: the email channel was used normally, at normal sizes, to normal recipients, while the volume went elsewhere.
The metadata has limits worth stating. It holds recipients, subjects, dates and attachment names and sizes, not message bodies or attachment contents. It covers the mailbox as synchronized to this machine, which may not include everything on the server. And the Deleted Items folder holds 892 items, which a full examination would read too, since a message sent and then deleted would not appear in Sent Items.
Module 10's email lesson reads the mail store itself.
The mailbox's rules complete the check:
Move newsletters 2025-08-14 Move to Archive
Flag client emails 2025-11-03 Flag for follow-up
Rules are read for their action and their creation date together. A rule created in the last weeks of employment, forwarding to an address outside the company, is the pattern to look for; a rule created months earlier that moves or flags mail is ordinary housekeeping.
Rules can also live only on the server and never appear in a local export, so a clean local result is confirmed against the server where it can be.
Two rules, both created months earlier, one moving newsletters to the archive and one flagging mail from a client's domain. Neither forwards anything. A forwarding rule to a personal address is one of the most common insider channels and one of the easiest to miss, because it sends mail silently after the rule is made. Its absence here is a finding, and lesson 0.6's finding listed it as one.
What Module 10 Builds On It
Every application that keeps its own recordModule 10 takes each application family in turn:
Module 10, in order
Module 10The browser lessons cover Chromium and Firefox in full, analysis across several browsers, and what private browsing does and does not write. The Electron and WebView2 lesson is increasingly important: Teams, Slack and many other desktop apps are built on browser engines and keep browser-style databases, so the browser methods apply to them.
The Windows Search database indexes file contents and email on the endpoint, and is sometimes the only record of a file that has been deleted. The email lesson reads mail stores on the endpoint, and the module ends on cloud application artifacts.
The application set on a typical workstation keeps growing, and each new one is a new place where a user's activity is written.
A chat app built on Electron keeps a local cache of messages; a cloud storage client keeps a database of synchronized files; a browser-based app may keep IndexedDB records in the browser profile. Module 10's approach is the same for all of them: find the profile folder, identify the databases, and read them against the timeline.
The habit to carry into it is to treat each application's data as a witness with its own scope. List every application and every profile before reading any of them, place what each holds on the case timeline rather than reading it alone, and size what a record proves to what it actually stores.
Practice
What do the browsers and the mailbox add? prepared evidenceThe prepared evidence below is the insider's Firefox search bar history, Chrome's downloads and the last week of sent mail. Separate the searches that bear on the case, find the download that set up the channel, and decide what the mailbox rules out.
If you have Firefox, close it and open formhistory.sqlite from your own profile with any SQLite browser. Read the values and their first and last used times, and notice how much of your own intent over the past months a single table holds.
Applications keep their own records in the user's profile, and in an insider case they often hold the intent. The second browser held five searches that sit between the channel's setup and its first use, asking how to do exactly what the other records show was then done.
Chrome held the download that set the channel up. The mailbox showed a channel that was available and not used, which the report states as a finding.
Across the two cases, the ten lessons read the same few machines from ten directions: what ran, who was there, how the attacker got in and moved, what was left to run again, what left, the filesystem beneath it all, what was hidden, the timeline, and the applications. Each direction is one module.
This is the last lesson of Module 0. Each of the ten lessons is the ground for one module of the course, and every one of them is read again, in depth, from Module 1 onward.
Resources
Microsoft revises these pages; where a page and this lesson disagree, the page is newer.