Lesson Report:
## Title: OSINT Media Verification Workshop — Territory vs. Time, Reverse Image Search, and Propaganda Framing
**Synopsis (2–3 sentences):**
This class session focused on helping student groups finish and refine their verification reports by clearly separating (1) the broader political/news claim from (2) what the image/video itself is actually evidencing. The instructor led a group-by-group review emphasizing *territory assessment* (geolocation plausibility/verification), then introduced how *temporal assessment* (when the media was created/first posted) can either matter a lot or not at all depending on the claim. The class ended with instructions to continue with reverse image search and prepare to move into deeper verification (especially of text) and political/propaganda analysis in the next sessions.
—
## Attendance
– **Students explicitly mentioned absent:** **1**
– Ruslan (Group 8) was noted as not present (“Ruslan is not here, right?â€�).
– Other groups were missing or delayed in the shared doc at times, but were not explicitly confirmed absent (e.g., Group 3 report not initially visible; Group 8 presentation did not occur).
—
## Topics Covered (Chronological, with detailed activity flow)
### 1) Session Goals, Week Roadmap, and Report Completion Focus
– Instructor opened by noting most students had completed required work and set the objective for the day:
– **Finish group reports** and use them to articulate the **reasoning behind propaganda/messaging** (whether verified or false).
– Preview of upcoming class: **Wednesday will focus on verifying text**, framed as *more challenging* than image/video verification.
– Instructor re-shared the **Google Doc link** via Zoom chat for students who lost access and explained the process:
– Go **group-by-group**, with **one representative** presenting:
– The **claim** made by the media
– The **media itself** (if not embedded)
– Their **territory assessment**: “Can we tell with certainty where this photo/video was taken?â€�
—
### 2) Group 1 Presentation — AI-altered Reporter Video, Olympics Claim (Milan Olympic Village)
**Activity/Case:** Pro-Kremlin Telegram video allegedly showing Ukrainian athletes placed far away in Olympic housing due to “bad behavior.�
– **Claim presented:** Ukrainian team housed “as far as possibleâ€� because Olympic administration learned from alleged past behavior (referencing Paris).
– **Media characteristics:**
– Uses footage of a **CBS News reporter**; later the reporter disappears and the video becomes images of athletes with an **AI voiceover**.
– **Territory assessment outcome:**
– Group concluded it was indeed filmed in the **Olympic Village in Milan** and “easily identifiable.â€�
– **Instructor feedback & methods highlighted:**
– Geolocation strategies:
– **Reverse image search**
– **Google Maps / Google Images** comparisons to known images of the Milan Olympic Village
– Instructor emphasized an important concept: sometimes **territory verification may not matter much** to the final authenticity determination (foreshadowing later work on overall verification and messaging).
—
### 3) Group 2 Presentation — China/Taiwan Tensions Video Using Archived Footage
**Activity/Case:** Video portraying China as preparing for war by crossing the median line in the Taiwan Strait; published by “Times Now� (Indian outlet).
– **Claim presented:** China crossing the median line; aggressive posture; video framed as intimidation of Taiwan. Title mentions **Trump** to provoke reactions/engagement.
– **Source/reliability notes by group:**
– Group researched outlet and suggested **fast “breaking newsâ€� style** may increase misinformation risk; noted fact-checkers/analysts have flagged issues.
– **Territory assessment:**
– Location broadly plausible (median line / Taiwan Strait context), but:
– **No identifiable geographic landmarks** in most clips; cannot pinpoint exact location.
– **Temporal discovery (key):**
– Using **Yandex reverse image search** on screenshots, group found similar frames appearing across **2015, 2021, 2022** → suggests **compiled/archival footage** rather than a single current event.
– **Instructor feedback & teaching points:**
– Reinforced scope: students are **not verifying the entire geopolitical claim** (“Is China escalating right now?â€�) but verifying whether **the footage matches the claimed context/time**.
– Highlighted a missed opportunity:
– Even without landmarks, students can verify **equipment identity** (planes, ships, missile systems).
– If hardware doesn’t match claimed actor/time (e.g., not Chinese equipment, or phased-out models), it can disqualify or contextualize the footage.
– Instructor framed why editing choices matter: combining bombastic military clips with a political claim aims to make viewers feel it represents “now.â€�
—
### 4) Group 3 Presentation — Iran Domestic Protests Video (Ekbatan Town, Tehran)
**Activity/Case:** Instagram video claiming protests in **Ekbatan** (Tehran) on **February 10**.
– **Claim presented:** Domestic protests occurring in Ekbatan town in Tehran; date asserted in caption.
– **Territory assessment (group):**
– Buildings resemble Ekbatan → **plausible**, but not fully confirmed due to lack of unique landmarks/street-level confirmation.
– Tools used: **ChatGPT** and **Picarta** (AI geolocation).
– **Instructor feedback & methodological emphasis:**
– AI tools can be a **starting point**, not the final authority—Picarta can be inconsistent (“off by whole continentsâ€� in instructor testing).
– Recommended OSINT next steps:
– Use **Google Maps** (satellite + **Street View**)
– Look for **signs**, **cars**, **license plates**, architectural features, vegetation, etc.
– Instructor deferred deeper **time/source verification** to later, keeping focus on territory assessment for the moment.
—
### 5) Group 4 Presentation — Russia/Ukraine War Damage Photo (Indoor Power Plant Setting)
**Activity/Case:** Photo (indoor, generic industrial setting) claimed to show **damaged thermal power plant** from Russian attacks; reposted with different timing.
– **Claim presented (as discussed):** Image used to imply Russian war crimes / attack on civilian infrastructure; request support for Ukraine framing.
– **Group approach:**
– **Reverse image search** found reposts on **The Guardian** and **Radio Free Europe/Radio Svoboda** (logo visible).
– **Instructor feedback (core lesson: authority vs verification):**
– “Triangulatingâ€� via reputable outlets is a start, but not sufficient: multiple sources can still be wrong; avoid relying solely on authority.
– Because it’s **indoors** and visually generic, geolocation is inherently difficult; instructor validated challenge but pushed for more rigor.
– **Key timeline issue discovered in discussion:**
– Group suggested post is from **2026**, but the damage alleged occurred in **2022** → territory and time are linked here.
– **Instructor recommended next steps:**
– Find when the image was **first posted/first published**.
– If photo predates alleged event (e.g., 2020), that would strongly contradict.
– If first appearance aligns with a known strike date, cross-reference attacks by date to narrow candidate power plants/cities.
—
### 6) Group 5 Presentation — Epstein Conspiracy Collage, AI-Generated Images, SynthID
**Activity/Case:** X (Twitter) post claiming NYC mayor **Zohran Mamdani** might be Jeffrey Epstein’s biological son; collage images alleged as “evidence.�
– **Claim presented:** Conspiracy narrative based on purported photos and tenuous social connections.
– **Group’s investigative findings:**
– Found **Reuters fact-check** stating many images **AI-generated**; mention of **Google SynthID** detection of AI watermark.
– Group also traced a referenced **email** in “Epstein filesâ€� and summarized Reuters’ reasoning that the email context does not support the conspiracy implication (Epstein likely not present at the event).
– Noted Mamdani’s mother (Mira Nair) reportedly told Reuters the images were **fabricated**.
– **Territory assessment conclusion:**
– Geolocation not possible because images are **AI-generated** / lack landmarks / no metadata.
– **Instructor feedback & skill-building:**
– Highlighted value of **publicly accessible primary sources** (ability to check files directly rather than relying on hearsay).
– Taught **SynthID concept**: invisible watermark indicating content generated/edited with Google’s systems (e.g., Gemini).
– Important correction: group relied on Reuters rather than testing themselves. Instructor requested students **run SynthID checks directly** when possible to strengthen the report (independent verification).
—
### 7) Group 6 Presentation — Israel Bombing Southern Lebanon Video (Context/Date vs Reality) + Metadata Misuse
**Activity/Case:** Instagram video posted Jan 30 claiming Israel warplanes bombing southern Lebanon “currently.�
– **Group reasoning:**
– Students noted Israel had conducted intense operations in late 2024–2025 but questioned whether the Jan 30 timestamp matched current events; suspected the **date/context** may be wrong even if footage is real.
– They attempted:
– Video “AI detectorâ€� (unnamed) → suggested video not fake
– **Metadata** extraction (but only from a screenshot)
– **Picarta** geolocation (conflicting with metadata-derived location)
– **Instructor feedback (critical technical correction):**
– Metadata from a **screenshot** reflects the device making the screenshot, not the original video’s capture metadata → cannot be used as original-source metadata.
– **Instructor-led OSINT alternative strategy:**
– Focus on available visual clues: **cars and license plates**
– Compare license plate format to known Lebanese plates (e.g., symbols/features), recognizing it’s not 100% conclusive but can be a useful clue.
– **Concept reinforced:**
– The class is verifying whether “this video shows X at time Y,â€� not proving the entire geopolitical situation.
—
### 8) Group 7 Presentation — Misattributed Mosque Destruction (Syria/Aleppo 2013)
**Activity/Case:** Video/posts claiming Kyrgyz military destroyed a mosque in a Kyrgyzstan–Tajikistan border conflict; group found it was actually Aleppo.
– **Claim presented:** Kyrgyz military destroyed a mosque (in border conflict).
– **Group finding:**
– The mosque is a **well-known site in Aleppo**, UNESCO-associated; destruction linked to conflict/terrorist organizations around **2013** (ISIS referenced).
– **Instructor response:**
– Acknowledged it’s “open-and-shutâ€� because the case is already widely investigated.
– Instructional note: for skill-building, students should ideally choose media that hasn’t already been fully resolved by prior investigators, otherwise they are only retracing steps.
—
### 9) Group 9 Presentation — Kyrgyzstan Political Photo in a Billiard Club (Hard Indoor Geolocation)
**Activity/Case:** Photo allegedly showing Komtibek Tashiev with politicians who signed a supportive letter; conflict with Tashiev’s claim he didn’t know them.
– **Claim presented:** Photo implies Tashiev is acquainted with letter signatories; used amid political dispute and dismissals/appointments.
– **Territory assessment challenge:**
– Indoor “genericâ€� environment (billiard club) makes geolocation extremely difficult.
– **Group approach:**
– Used **Picarta** (exact GPS behind paywall)
– Hypothesized match to a specific venue: **Billiard club “Kondorâ€�** based on visual similarity (lighting/hall appearance) found via Google Maps/internet photos.
– **Instructor feedback:**
– Encouraged adding comparison photos into the doc (strong OSINT practice: explicit visual matching evidence).
– Noted verification may remain impossible, but comparison-based narrowing is valuable.
—
### 10) Whole-Class Synthesis + Instructions for Next Step (Temporal Assessment + Reverse Image Search)
**Key synthesis skill emphasized:**
– Separate:
– The broader **news/political claim**
– From what the **image/video itself can verify**
– Introduced **temporal assessment logic** with examples:
– **Tashiev photo**: If found to be 10 years old, it **does not necessarily** disprove the claim “he knew/didn’t know these people,â€� because the claim is about acquaintance, not a time-specific event.
– **Lebanon bombing video**: If found to be from 2024, it **does** disprove the claim it shows bombing on Jan 30, 2026 (date/context mismatch).
**Assigned in-class task (started but deferred):**
– Spend ~10 minutes doing **reverse image search** (Google + Yandex preferred).
– For videos: take multiple screenshots at key moments and search each frame.
– Determine: if earlier appearances are found, does that **change the claim’s truth value** depending on what the claim asserts?
—
### 11) Q&A: Using Articles That Say “Photo Is Not From This Event� + Instructor Guidance
– Student asked about cases where articles explicitly say the photo is not from the described event (stock/illustrative photo).
– Instructor response:
– Such claims from outlets are useful signals but not definitive; outlets can be wrong or biased.
– Best practice: **independent verification** whenever possible to strengthen students’ own assessment.
—
### 12) Wrap-Up / Next Class Plan
– Due to time, the reverse image search activity was postponed to **start on Wednesday**.
– Future workflow:
1) finish temporal assessment / reverse image search
2) move into the final section: **political analysis / propaganda messaging**
– Class ended ~2 minutes early.
—
## Actionable Items (Short bullets, organized by urgency)
### High Urgency (Before Wednesday / Next Class Start)
– **All groups:** Complete/continue **reverse image search** (Google + Yandex).
– Videos: take **multiple screenshots** from distinct moments and search each.
– **All groups:** Add results to the shared Google Doc in a way that supports reconstruction: links, screenshots, “first seenâ€� dates, and brief reasoning.
– **Group 5 (Epstein collage):** Actually **upload images to Gemini/SynthID check** (don’t rely only on Reuters’ conclusion) and record outcomes.
– **Group 6 (Lebanon bombing video):** Stop using screenshot metadata as “original metadata.â€� Re-do verification using **frame reverse search** and **license plate comparison** if clearer frames exist.
– **Group 9 (billiard club photo):** Insert **visual comparison evidence** (photos of “Kondorâ€� or candidate venues) into the doc.
### Medium Urgency (Improve Rigor / Strengthen Reports)
– **Groups relying on “reputable outlet reposted itâ€� (e.g., Group 4):** Add independent verification steps beyond authority (first-post date tracing; contextual strike timelines; additional corroborating visuals).
– **Groups using AI geolocation tools (Picarta/ChatGPT):** Treat AI output as a lead only; corroborate using **Street View, satellite, signage, vehicles, license plates, architecture**.
– **Group 7:** Select a **new, unresolved** media item if possible (current case is already widely investigated), to demonstrate original verification work.
### Upcoming / Instructor Follow-Up
– Instructor to begin Wednesday with the deferred activity: **temporal assessment + reverse image search logic** and then transition to **political/propaganda analysis** section of reports.
– Reminder: Wednesday focus will shift toward **text verification** (explicitly described as more challenging than images/videos).
Homework Instructions:
NO HOMEWORK — The instructor only directs in-class work (“spend the next 10 minutes beginning that [reverse image] search�) and says the unfinished activity will be done in the next meeting (“We’ll be starting… on Wednesday’s class with this activity… We’re going to be finishing up the reports�), without assigning any out-of-class task to complete before Wednesday.