Beyond the Blackout: Documenting Atrocities in Tigray’s Forgotten War

Beyond the Blackout: Documenting Atrocities in Tigray's Forgotten War

Local journalists and healthcare professionals risked their lives to collect fragile digital and analog evidence of the 21st century’s deadliest conflict.

Team

Reading Time: 5min

Background

In response to the 2020-22 war in the Tigray region of northern Ethiopia and the government-imposed communications blackout and systematic blockade that obscured atrocities from external scrutiny, Tigrayan civil society actors launched a coordinated initiative to preserve digital evidence of war crimes and human rights violations. 

Although formal restrictions are technically lifted, the extensive destruction of telecommunications and power infrastructure, combined with the incomplete implementation of the Pretoria Agreement, has created a sustained, de facto blackout, leaving a significant portion of the population digitally isolated. To counter state censorship and mitigate the risk of evidentiary loss and historical erasure, the initiative employs a deliberate, multi-phased strategy that includes the collection, exfiltration, and archiving of fragile digital materials.

Conceived as both a protective measure and an act of resistance, the project advances three long-term objectives: (1) building a credible evidentiary base for future accountability and transitional justice processes; (2) protecting the historical record of affected communities against denial, manipulation, or revision; and (3) enabling ongoing research, digital forensics, and human rights analysis by scholars, practitioners, and advocates.

Context

The conflict that erupted in November 2020 in Ethiopia’s Tigray region rapidly devolved into one of the deadliest wars of the 21st century. Credible estimates indicate up to 600,000 civilian deaths, the vast majority not through direct combat but through starvation, preventable disease, and the collapse of essential services. More than 2 million people were internally displaced, nearly one million fled as refugees to neighboring countries, millions were subjected to mass starvation, and more than 100,000 were victims of forms of sexual violence.

From the conflict’s outset, the Ethiopian government, in coordination with the Eritrean Defense Forces (EDF) and Amhara regional forces, imposed a systematic siege that functioned as a strategy of collective punishment. Essential services, including telecommunications, banking, electricity, fuel, and commercial supply chains, were deliberately severed. This weaponization of deprivation was compounded by the routine obstruction of aid convoys, the suspension or expulsion of relief organizations, and the denial of access to independent media and the expulsion of journalists. Bureaucratic restrictions and relentless checkpoint inspections deliberately impeded the delivery of food, medicine, and other life-sustaining resources. These integrated measures cut off vital remittances, halted commerce, prevented civilians from accessing critical care, and enforced a state of utter isolation. The siege was reinforced by a state-imposed information blackout, comprising the longest consecutive internet shutdown on record and coordinated digital disinformation and hate speech campaigns, which successfully obscured the scale of violations and shielded perpetrators from international scrutiny. All parties to the conflict have been accused of war crimes.

The severity of the atrocities has led legal scholars to conclude the conduct of the war may meet the legal threshold for genocide. The Tigrayan community has consistently described the campaign as a deliberate genocide. The pathway to accountability was systematically blocked: the Ethiopian government denied access to the UN’s International Commission of Human Rights Experts on Ethiopia (ICHREE), and the Commission’s mandate was ultimately dissolved in October 2023.  This outcome left grave crimes unexamined and institutionalized impunity, especially since institutions implicated in the violations retained positions of authority over the national transitional justice process.

 

It is within this systemic accountability vacuum that Tigrayan civil society actors mobilized to safeguard digital evidence of atrocities. As one civilian journalist stated, “The most precious asset is information now, not gold, and it needs to be protected.” This atmosphere of fear, censorship, and violence created an urgent need for an independent, community-based effort to archive and preserve evidence of the atrocities being committed. 

In late 2022, a member of our team traveled to Tigray to meet with a healthcare professional and a civilian journalist who had each spent more than a year risking their lives to document atrocities using whatever tools were available to them. Both relied heavily on their mobile phones to capture photos, videos, and testimonies, assuming the devices themselves were a safe form of storage. When asked about their preservation plans, they didn’t have one, they had been too focused on survival to consider what might happen if a screen cracked, a battery failed, or a phone was confiscated at a checkpoint. When digital capture wasn’t possible, they turned to paper and pen, methods even more vulnerable to loss. 

Their stories were not unique. They reflected a stark and pervasive gap between the extraordinary courage of frontline documentarians and the technical knowledge required to ensure that evidence survives long enough to reach a courtroom. That gap is what gave rise to this initiative. In anticipation of denial, erasure, and the consolidation of state narratives, civil society actors began the systematic collection, exfiltration, and preservation of digital and analog evidence.

Framework

The Challenge

The team’s methodology was forged by the extreme hostility of the operating environment. Working with a small team and limited financial resources, every decision was a direct response to a specific set of constraints designed to prevent evidence from leaving the region. The key challenges included:

A Total Information Blackout
This was not just a matter of slow internet. The team’s data collection took place during the longest consecutive internet shutdown in history, creating a total information blackout that made digital transfer of evidence impossible. This forced a complete reliance on offline-first solutions. As partial connectivity resumed, the aggregation and transport of data began, but the team faced a new set of challenges. The state-controlled ISP actively blocked access to essential security tools, including the Proton suite, specific VPNs, and the Bitwarden password manager. Furthermore, in areas with connectivity like Mekelle, the connection speed was so low that it could not handle multimedia datasets; attempts to upload video evidence (over 1,500 files totaling 450 MB) would consistently time out.

Extreme Physical and Personal Risk
Moving data was a life-threatening activity, exposing the team to different risks across the operational chain: 

Local Data Collectors: Their exposure centered on the immediate liability associated with the initial acquisition and transient possession of sensitive material. This vulnerability created a risk of surveillance and arbitrary detention during the collection process if caught.

Central Custodians: These managers, responsible for local data aggregation at a central site, faced a significant, persistent liability. The data was stored temporarily in private residences, meaning there was no secure facility for staging the evidence. Every piece of evidence constituted a constant threat of discovery via raid or accidental operational exposure.

Transporters: Individuals tasked with the physical movement of data for exfiltration encountered the most acute threat. They risked being stopped and searched at checkpoints, where finding a hard drive could lead to imprisonment on fabricated charges of espionage and terrorism. The risk only subsided once the evidence physically left the country.

Severe Resource Scarcity
The war created a severe scarcity of basic technology. Laptops and computers had been widely destroyed or looted, and the few devices that remained were often older models, unable to receive software upgrades and lacking sufficient storage. Key items like external SSDs were not just scarce in Tigray but unavailable for reliable purchase in Ethiopia as a whole. While a secondary market existed, it was too untrustworthy due to the high risk of malware or corrupted hardware. To mitigate this, the team adopted a strict procurement strategy: all equipment was purchased new in the United States. This approach addressed the dual challenges of trust – ensuring devices were not tampered with – and logistics, as it was less of a red flag than a bulk purchase in the region. A team member then transported all hardware personally in their carry-on luggage, maintaining a secure chain of custody and bypassing the risks of confiscation or damage associated with checked baggage.

Ensuring Data Integrity Across a Human Chain
With no digital transfer options, data had to be physically transported via a “human chain of custody.” This method was not only a security challenge but also a significant logistical gamble. The entire process was dependent on international travel from the US to Tigray and back – a journey with no direct flights, requiring layovers in Addis Ababa. As a result, the mission was constantly at the mercy of potential flight cancellations, which could trap the data courier and halt the extraction process entirely. The team, therefore, had to ensure not only that the data was protected from tampering, corruption, or loss, but that its perilous physical journey could be completed at all.

20210816_101955

The Prototype

The project’s methodology was built around a three-phase lifecycle for each piece of evidence: secure capture, long-term preservation, and availability for future analysis – echoing our Starling Framework. This framework was designed to be resilient in a low-tech, high-risk environment.

Secure Capture of Field Data

The data capture strategy prioritized security and ease of use for contributors on the ground. The team recognized that complex security protocols could increase the risk of human error, so user-friendly tools were selected. This approach ensured that journalists and civil society members could focus on documenting events without being hampered by technical complexity. All equipment was successfully transported to Tigray, and the local team received training in encryption, data security best practices, and storage resilience through redundancy.

A 3-2-1 Archival Strategy for Preservation

For long-term storage, the team adopted the 3-2-1 archival best practice: maintaining three copies of the data, on two different types of media, with at least one copy stored off-site. This strategy was chosen to mitigate the risks of data loss due to hardware failure, corruption, or physical seizure of any single data store. This distributed approach ensures redundancy and the long-term availability of the evidence.

Technology

The project’s technical methodology was designed for an environment of extreme hostility and minimal resources. Every tool and process was chosen to prioritize user safety, data integrity, and resilience, following the Starling Framework principles of Capture, Store, and Verify.

However, it’s critical to define the boundaries of this workflow. The cryptographic integrity measures begin the moment a contributor’s data is ingested into our central MacBook in Tigray. This starting point was a necessary adaptation to the realities of sourcing; because some data was ingested off-the-record from sensitive sources via direct drive copies or downloads, formal handoffs using digital signatures or email were impossible. The journey of the data before this point – from the original capture device to the contributor – is not cryptographically documented in this process. Our chain of custody therefore covers a “known universe” ranging from ingestion to final archival storage.

This can be visualized as follows:

  • Phase 1 (Undocumented): Event Capture → Initial Transfer to Contributor Storage
  • Phase 2 (Documented & Verified): Ingestion → Hashing → PGP Signing → Timestamping
  • Phase 3 (Out of Band): Phase 2 Files → Transport → Copy to Redundant Storage

To verify, retrieve files from redundant storage, decrypt files to actual files, verify against Phase 2 records.

Capture and Initial Processing

The capture phase focused on safely ingesting data from contributors and securing it for transport. The primary challenge was the trade-off between ideal security and the practical usability of tools for non-technical journalists operating under extreme stress.

“Just a Bunch of Drives” (JBOD in archiving jargon) in the team’s carry-on luggage

Choice of Field Hardware
The core field kit included a new 2023 16-inch MacBook Pro and multiple Samsung T7 Shield SSDs. The MacBook Pro was specifically requested by a core contributor who was already familiar with the product from their video journalism work, minimizing the learning curve. It was chosen for its powerful graphic rendering and storage capabilities, which were essential for handling high-resolution media. Procuring a new device also ensured it would be compatible with software upgrades and other devices over a multi-year horizon. The Samsung T7 Shield SSDs were selected for their durability and built-in AES 256-bit hardware encryption, providing a crucial layer of security during transit.

Use of Encryption and Usability
We made a deliberate choice to prioritize simple, reliable encryption over more complex, error-prone alternatives. For the laptop, we relied on the MacBook’s native FileVault for “set-and-forget” full-disk encryption, which minimized the risk of human error under duress.

For the external SSDs, the team used symmetric encryption, storing the shared passwords in Bitwarden. While asymmetric encryption was considered (encrypting files separately for each person), it was rejected because it would have made key management too challenging across stakeholders. The symmetric approach was also more communal; since everyone had the same key, it made the team feel more comfortable, a deliberate trade-off for a more manageable and inclusive security posture.

Broad Use of Secure Communications
All team coordination and sensitive communications were conducted via both Signal Messenger as well as WhatsApp (the most widely-used messenger locally), leveraging its strong, open-source, end-to-end encryption.

Store

The storage strategy was built on the archival best practice of 3-2-1 (three copies, on two different media, with one off-site) to ensure redundancy and resilience against data loss. The choice of storage solutions was the result of a formal evaluation that prioritized security and censorship resistance.

Criteria

Dropbox / Google Drive

Proton Drive / Tresorit

AWS S3 / B2

Usability

High

Medium

Low

Privacy

Low (No E2EE)

High (E2EE by default)

Medium (User-configured)

Censorship Risk

High

Low (Swiss Jurisdiction)

High (US Jurisdiction)

Based on this analysis, the following three-tiered storage model was implemented:

  1. Primary Cloud Storage: Proton Drive was selected for its default end-to-end encryption and its legal jurisdiction in Switzerland, which offers strong privacy protections. This served as the primary off-site copy.
  2. Local Device Storage: Two separate Mac laptops (one in Tigray, one in the US) held complete, encrypted copies of the archive.

External Physical Media: A dedicated external hard drive stored the third copy, providing an additional layer of media diversity.

Verify Archiving or Publishing

The verification phase was designed to create a transparent, auditable record proving that the collected data was not tampered with after ingestion. This process established a cryptographic chain of custody for our “known universe” of data – from the moment we received it to its final archival.

Step 1:Hashing for Integrity
The first step was to create a unique cryptographic “fingerprint” for every single file. Using the CLI, we generated a complete manifest of SHA-256 hashes. This provided a baseline record of the dataset’s exact state at the time of ingestion.

$ find . -type f -exec shasum -a 256 {} \; > TigrayData_V1_fingerprints.txt

Step 2: Signing for Authenticity
The resulting *_fingerprints.txt manifest was then cryptographically signed using an organizational PGP key. This signature acted as a digital seal, proving that the manifest itself is authentic and has not been altered since it was created by our team. 

The plaintext signature remains publicly verifiable with the .sig file: anyone can verify that the signature was applied by this specific PGP key – without requiring decryption.

$ gpg --import keys/public.asc
key 55082AE8487FB65C: "Tigray Archive <tigrayarchive@proton.me>"
Total number processed: 1

$ gpg --verify signatures/TigrayData_V1_fingerprints.txt.sig \
data/TigrayData_V1_fingerprints.txt
TigrayData_V1_fingerprints.txt
  ⎿ gpg: Signature made Wed Oct 15 22:44:18 2025 CEST
    using RSA key 9945877FA0290DB20547B81C55082AE8487FB65C
    Good signature from "Tigray Archive <tigrayarchive@proton.me>"
(...)

 

Step 3: Timestamping for Proof of Existence
To establish an immutable, publicly verifiable record of when this dataset existed, the team used OpenTimestamps—a service that anchors cryptographic proofs to the Bitcoin blockchain.

During active conflict operations, the team timestamped the encrypted signed manifest (.sig.gpg file). This served dual purposes: it created a blockchain-anchored proof that a specific sealed dataset existed at a specific time, while keeping the file list confidential to protect sources, operational security, and the safety of individuals still documenting evidence under threat.

For public release and independent verification, the team is now publishing the unencrypted signed manifest with a new OpenTimestamps proof. This enables anyone – researchers, investigators, legal practitioners, journalists – to verify the exact file list and SHA-256 hashes without requiring decryption keys or trusting intermediaries.

$ cat TigrayData_V1_fingerprints.txt
a50d392fb51080c1052991e17cd1ac8b1d642fc36dc766a771a3ac2666773605 ./ttv jounalist/himemte shekore.mxf
a2a667af88b9f6443734011ade36f99c577f27f473bd263813ddf008cd500f62 ./ttv jounalist/emahoy tiemtu.MXF
...

 

Commits are signed:

commit 93b45d...13276897 (HEAD -> main, origin/main)
Signature made Sat Oct 25 09:05:30 2025 CEST
using RSA key B826CF0ED10B023C411879FD9B66FE79F29FFF9C

Good signature from "Tigray Archive <tigrayarchive@proton.me>" [ultimate]
Author: Basile Simon <tigrayarchive@proton.me>
Date:   Sat Oct 25 09:05:30 2025 +0200

    Initial commit

 

The companion GitHub repository to this case study, which contains the public cryptographic proofs that allow any third party to independently verify the integrity and provenance of the collected data, contains four key files that work together to establish provenance:

  • TigrayData_V1_fingerprints.txt – The complete manifest, listing the SHA-256 hash for every file in the dataset.
  • TigrayData_V1_fingerprints.txt.sig – A GPG signature of the manifest file, creating a tamper-proof seal.
  • TigrayData_V1_fingerprints.txt.sig.ots – An OpenTimestamps proof, which anchors the signature to the Bitcoin blockchain to prove when the manifest was signed.
  • pubkey.asc – The Tigray Archive public GPG key used to create the signature. A user must import this key to verify the manifest’s authenticity.

Learnings

Adapting to a Low-Tech, High-Risk Environment

This project underscored the necessity of adapting security protocols to the realities of the operating environment. The team learned that an emphasis on offline-first solutions was more effective than complex digital tools that could fail due to infrastructure limitations. This reliance on physical transport, robust hardware, and user-friendly encryption was successful for nearly every aspect of the workflow, with the notable exception of timestamping, which remained a challenge due to the severe connectivity issues.

 

The Importance of Trust in Community-Based Archiving

Trust was the cornerstone of this initiative. In a conflict affecting a tight-knit community like Tigray, every decision—from selecting tools to identifying data sources and training local journalists—was made collectively. Nothing proceeded without the full consensus of community members.

This collaborative structure was possible because the initiative grew from years of voluntary work and credibility earned through lived experience. With fewer than six million people and minimal degrees of separation between families, everyone in Tigray is traceable within the social and political landscape. This closeness fostered trust but introduced risks absent with outside actors. Local sharing carries cultural and social sensitivities, such as the fear of bringing shame to families of survivors, particularly in cases of sexual violence. It also risks being misinterpreted: grassroots evidence collection can be viewed as operating outside official channels, raising suspicion that shared data might be commodified or used for personal gain.

In this environment, consensus-based decision-making and trust as the organizing principle were essential and enabled the work.

This collaborative approach was critical for the project’s success and the ethical stewardship of highly sensitive data. The project demonstrated that technology is merely a facilitator; the real work lies in building human networks of trust. Interestingly, diaspora and foreign entities were often trusted more than local journalists, as they were perceived to have greater influence in justice and accountability processes.

 

A Need for Accessible and Resilient Tools

The challenges faced in Tigray highlight a broader need for cheap, accessible, and open-source tools designed for evidence preservation in resource-constrained environments. The success of this project relied on commercially available, consumer-grade products that were simple to use. 

The project’s most significant technical challenge was balancing security best practices with the reality of the operating environment. While tools like VeraCrypt (for creating hidden, encrypted volumes) and TailsOS (for anonymous browsing) were considered, they were ultimately rejected. In a high-stress conflict zone with non-technical users, compounded by the fact that the local team’s second language was English, the complexity of these tools was deemed a liability; a single mistake in their use could lead to the permanent loss of critical data.

This led to a crucial and deliberate compromise. The team encountered errors when using OpenPGP to encrypt large individual files and folders, which would have provided the most granular, asset-level security. Instead of risking data corruption or overwhelming users with complex command-line workarounds, the team opted for simpler, device-level encryption using the native and user-friendly FileVault on the MacBooks. While technically less secure than encrypting each file individually, this approach provided robust, reliable protection for the entire dataset and, most importantly, minimized the risk of human error, which was identified as the greatest threat to the data.

Finally, the initial assumption that large, terabyte-sized drives would be optimal for storage provisioning proved incorrect. We initially provisioned drives ranging from 1TB to 4TB. However, optimal utilization dictated that drives with significantly smaller storage capacities, approximating 128GB, would have been more effective for field distribution. The rationale was twofold: the collected data was highly distributed, and local journalists required dedicated copies. Provisioning 1TB+ units resulted in low storage utilization and substantial empty space. The team ultimately accepted the compromise of larger drives, as deploying the necessary volume of 128GB units would have substantially increased the logistical burden and operational risk associated with transporting a higher equipment count.

 

User Experience as a Security Imperative

The project revealed that in a high-stakes field environment, user experience is a critical component of security. We initially considered deploying gold-standard security tools like TailsOS for anonymity and VeraCrypt for creating hidden encrypted containers – see point above.

However, in practice with non-technical journalists operating under extreme stress, these tools proved to be a liability.  A complex interface with multiple steps increased the likelihood of a user making a mistake that could either fail to protect the data or, worse, lead to its permanent loss, or trigger unnecessary operational redundancy, increasing costs and harming team moraleWe learned that the greatest immediate risk was not a sophisticated adversary cracking our encryption, but a well-meaning user accidentally locking themselves out of a hidden volume forever.

As a result, we actively chose simpler, more intuitive tools. Showing a journalist how to enable the native FileVault encryption on their MacBook – a one-time, set-and-forget process – was infinitely more effective and secure in practice than training them on a complex, multi-step process they would have to remember under duress. This experience demonstrated that the most secure tool is often the one that a user can reliably operate correctly every single time. Crucially, the tools had to be zero-cost solutions. Given that the operating environment in Tigray is largely cash-based and lacks widespread digital payment infrastructure, any solution requiring a subscription or online purchase was deemed inherently sub-optimal.

 

The Human Cost and Cognitive Burden

Beyond the technical and logistical complexities, the project illuminated the human toll of evidence preservation in conflict settings. Civil society actors bore a dual burden: conducting high-risk documentation under constant security threats while also carrying the weight of their own trauma as survivors or witnesses and engaging directly with other survivors of violence. This required navigating a delicate balance between the imperative to document and the ethical obligation to center survivor dignity and sensitivity. High-stress conditions, fractured infrastructure, and perpetual risk of exposure compounded the psychological strain. In this context, user error and burnout were not incidental operational risks but outcomes of the challenging environment itself.

For future efforts, two areas for improvement stand out. First, dedicated Mental Health and Psychosocial Support (MHPSS) resources should be integrated into operational design to address the cumulative psychological burden on frontline teams. Second, establishing core staffing structures with individuals more operationally removed from the conflict can help sustain continuity, reduce direct exposure of local actors, and ensure the work is not solely borne by those most at risk.

Recognizing psychosocial resilience as inseparable from technical security is central to the integrity and sustainability of evidence preservation in conflict environments. The most important insight was that resilience and trust proved as critical to data integrity as any protocol or hardware. Technical safeguards alone could not succeed without deliberate attention to human and psychosocial dimensions.

Archive & Appendix: Case Summaries of Documented Atrocities

The following summaries detail five incidents documented in this dataset. This appendix is intended to serve as a resource for researchers, investigators, and legal practitioners by providing foundational information and leads for further inquiry into alleged war crimes and crimes against humanity. All listed incidents have been subject to public reporting. Utmost care must be taken to protect the identity of sources when utilizing any related data.

Researchers should reach out to info@starlinglab.org to discuss access to the documentation.

Case 1: Debre Abay Monastery Massacre

Case 2: Togoga Massacre

Case 3: Teka Tesfay Massacre

Case 4: al-Nejashi Mosque Shelling and Looting

Case 5: Destruction and Looting of Atse Yohannes High School



Privacy Preference Center