ICC Forum

A Single Comment — Permalink

Comment on the Cyber Evidence Question: “To what extent can cyber evidence repositories, and digital and open-source evidence, facilitate the work of the OTP, and the ICC more generally?”

Protecting Against Deepfakes: How the ICC Can Ensure Trust in the Verification and Use of Open-Source Evidence

I. Introduction

Digital devices have allowed ordinary civilians to become on-the-ground investigative reporters in almost every region of the world. Contemporaneous uploads of footage featuring protestors being tear-gassed in Hong Kong or Syrians fleeing from chemical warfare have allowed viewers all over the world to see, hear, and emotionally feel human rights crises unfold in real-time. As a result, the amount of online evidence and information available to prosecutors and investigators through open-source channels has raised the possibility of reinvigorating—or even changing the narrative—in formerly stalled investigations. The Office of the Prosecutor (OTP) has already begun to utilize open-source information from social media and other online platforms to further investigations. In 2017, the ICC issued a first-of-its-kind warrant for the arrest of Mahmoud Al-Werfalli, a commander within the Libyan National Army, based on open-source videos.1 In the sphere of international criminal law and the fight against human rights abuses, the new and increasing use of open-source information has opened up the possibility for investigations to be better, safer, and more efficient than ever before.

Open-source refers to publicly available online materials, ranging from photographs to documents to audio and visual recordings.2 Such information, commonly housed on social media sites, is often posted by ordinary civilians and is available and growing at a prodigious rate.3 The benefits of open-source information are clear. As Professor Danielle Citron of Boston University School of Law testified to the U.S. House Permanent Select Committee on Intelligence, the use of video and audio recordings obtained through online public records has allowed us to “become firsthand witnesses to events, obviating the need to trust others’ accounts.”4 Content captured contemporaneously by any individual with a capable phone or device has the power to strengthen investigations in ways that documentation and evidence gathered at a later time by investigators plainly do not. Open-source data enables researchers to advance investigations from anywhere around the world rather than risk traveling to volatile regions where they may be severely restricted or put in danger. At the same time, the use of such data in the judicial system has the potential to make investigations safer by corroborating and giving strength to witness testimony or by lowering the necessity for vulnerable witnesses to testify in court in the first place.5 Rebecca Hamilton, formerly of the OTP, suggests open-source evidence has the ability to “democratize” investigations by giving power to local communities and their own ability to bring an end to the abuses they have suffered or witnessed first-hand.6

Expand

An increased reliance on and use of open-source information in the judicial system is not without consequences. A growing trust in the authenticity of content posted online has given bad actors the opportunity to alter open-source data to their advantage in order to convey a false or dangerously misleading narrative separate from reality. The term ‘deepfake’ is used to refer to digitally distorted content such as “videos generated via algorithms that make it look like a person said or did something she did not.”7 Deepfakes, originally popularized in the porn industry, are typically sets of data merged together.8 As Professor Citron articulated, “[c]reators of deepfakes count on us to rely on what our eyes and ears are telling us, and therein lies the danger.”9 As of October 2019, at least 14,678 videos considered to be deepfakes exist online and are free for anyone to access—and to erroneously interpret as true—around the world.10

Addressing the problems that go into verifying and authenticating open-source evidence to protect against the rising prevalence of deepfakes is particularly difficult given the ICC’s lack of financial and personnel resources available to tackle the extraordinary amount of open-source information that continues to multiply every day. Thus, this comment argues that the ICC must develop partnerships with third-party organizations already employing trained data scientists with the forensic knowledge to equip the ICC with resources to publicly verify open-source evidence. A failure to adapt and evolve in the digital age could cause the ICC to risk losing relevance and public trust during a time when the collection of key evidence in human rights investigations may be more critical than ever.

This comment does not touch on the complex evidentiary rules governing the Court that likely pose separate challenges to the use of open-source evidence, the difficulties regarding authenticating the chain-of-command when using publicly-sourced evidence, or the use of user-generated applications that seek to protect the anonymity of those who provide open-source evidence in the first place. This comment does, however, focus on the need for the ICC to hold itself to a higher standard as a permanent international tribunal and argues that the Court must establish a blanket verification process for all open-source evidence. A strong and neutral authentication process will safeguard the validity of the OTP’s investigations, the rights of the accused whose freedoms may hinge on the soundness of such evidence, and the ICC as a whole. To do so, this comment will discuss the verification approaches of other non-state actors that offer paradigmatic examples for the ICC and the OTP to learn from. This comment then seeks to expand on arguments and recommendations by other qualified experts in the field.

II. Paradigmatic Verification Models by Non-State Actors

The consequences of failing to authenticate and verify open-source information are obvious and severe. False convictions contingent on doctored evidence present a human rights challenge in itself; allowing a doctored video to enter through the cracks of the judicial system risks undermining trust in prosecutorial investigations at a time when global confidence in the media is critically low.11 As of now, the OTP appears to engage primarily in simple verification schemes focused only on substantiating the time, date, and location of publicly provided evidence.12 But to evolve in an era where bad actors are constantly scheming for ways to outsmart and outmaneuver prosecutors and academics with advanced digital capabilities and new technology, the ICC must do more to advance its verification capabilities beyond being able to identify obvious open-source alterations. Lessons can still be taught from Reuters’ inadvertent use of a digitally altered photo in 2007 of airstrikes during the Lebanon war that appeared to show more smoke rising from military action than in reality.13 The New York Times has also had to publicly confront the aftermath of a similar mistake when they published a photograph distributed by Sepah News, Iran’s state media site, of an Iranian missile test in 2008 that was discovered to have likely been digitally altered to look as if there was an additional missile (note that there is no confirmation of official Iranian government involvement in the doctored photograph).14

Amnesty International has led the way in training volunteers with verification skills through their creation of the Digital Verification Corps (DVC).15 The DVC was created to specifically tackle the authenticity challenges faced by an overwhelming amount of open-source evidence. With roughly 120 student volunteers installed in at least six universities worldwide and growing, the DVC continues to train “highly-qualified video verifiers” in an effort to assimilate digital verification with critical human rights investigations.16 Sam Dubberley, the manager of the DVC, stated:

When we are not on the ground to witness these events as they happen, the onus is on us to prove beyond a shadow of a doubt that they really did happen.17

In 2018, the DVC single-handedly took the lead in verifying video footage that appeared to show soldiers shooting two women and children in the head, despite the Cameroonian government’s denial and resolute claim that the video was a deepfake.18 As a result of the DVC’s analysis, the Cameroonian government finally arrested the soldiers and acknowledged the executions.19 With a growing presence around the world of trained and passionate students, the DVC stands at the ready to analyze and verify open-source content in varying languages and time-zones.20

The UC Berkeley Human Rights Investigations Lab has also played a leading role in digital verification by training and combining resources with the DVC.21 The Investigations Lab has orchestrated and hosted workshops to engage and train students with technology at the forefront of investigative research with the goal of advancing investigative abilities to tackle human rights abuses with reliable open-source information.22 At the same time, the UC Berkeley Human Rights Center has already engaged with the OTP through similar workshops aimed at educating investigators on how to use and verify open-source information in order to overcome evidentiary thresholds.23

Bellingcat is a different kind of open-source investigative site, started by an international blogger, that integrates and equips journalists with new technologies and publicly sourced information.24 Bellingcat prioritizes collaborating with other organizations and contributors in order to grow its network and spread knowledge and investigative skills on how to better use open-source evidence in the investigative context.25 The group has already begun to assist the ICC in the Court’s understanding and use of open-source information.26 Finally, the Atlantic Council’s Digital Forensic Research Lab is another prominent example of a leading nonpartisan organization with a mission to break down misinformation on platforms that host open-source information.27 Housed with digital forensic analysts, the Digital Forensic Research Lab tracks international events and publishes reports that seek to dissect open-source methodologies and shed light on distortion campaigns happening all over the world.28

III. Recommendations

The ICC must adopt a coordinated solution to an international challenge that knows no borders and has no boundaries. Regardless of the inculpatory or exculpatory value of certain open-source evidence the ICC seeks to use in an investigation, the ICC —and in turn the OTP —must use the same consistent and public verification process for authenticating open-source material. A consistent authentication process will go a long way in fostering trust in the ICC’s use of such open-sourced data in the first place. This is particularly important in cases where an investigation hinges on a piece of video or audio evidence obtained on an open platform, as the Court must be particularly mindful in order to protect all parties involved in proceedings before the tribunal—including the accused. As commentators have noted, the use of open-source evidence carries with it many concerns for defendants, such as:

[I]ts proclivity to convey a prosecution-biased picture of the crimes and to exacerbate the inequality of arms that already disadvantages defendants in international criminal proceedings.29

When the burden is often shifted to the defendant to prove reliability—or lack thereof, the Court should do everything it can to make sure proceedings are fair and defendants have the same ability to use open-source evidence as the prosecutors. Investigations particularly dependent on open-source evidence should have an even higher impetus to prove reliability through an immediate and public verification process.

Second, the ICC must establish concrete partnerships with third-party investigative bodies. Experts like Hamilton have acknowledged the need for the OTP to engage new players in understanding, using, and verifying the massive amounts of data relevant to the Prosecutor’s investigations and proceedings before the ICC.30 The DVC’s unique utilization of university students is a model example of how to overcome budget challenges; modern day academics are being outrun by individuals with forensic knowledge and the ICC should take the initiative to learn from—and train—students with the digital know-how to quickly and accurately verify and spot digitally altered open-source data. Involving students in the ICC’s investigative proceedings has the dual purpose of both advancing the ICC’s ability to function and training the next generation of human rights defenders. The ICC should furthermore require the OTP to communicate with experts throughout the investigative process, ensuring such experts are available to speak to the truthfulness and verification of open-source evidence when needed.

Lastly, the ICC and the OTP must be forthright in acknowledging the consequences that come with too heavy of a reliance on open-source evidence. As technology makes it even easier to create deepfakes, hostile state and nonstate actors may be incentivized to forge open-source evidence as leverage. At the same time, a growing awareness of the prevalence of deepfake content might cast doubt on open-source evidence in a perverse way; furthering what Professor Citron calls the “liar’s dividend,” where liars try to avoid culpability by twisting the narrative and claiming a real video is in fact fake. 31 As aforementioned, the Cameroonian government tried to get away with doing just that.32 Calls claiming open-source evidence is ‘fake news’ may discredit the ICC’s efforts to integrate open-source evidence into formal proceedings, further justifying the need for a coordinated and neutral verification system to shield against attempts to discredit the use of open-source information.

IV. Conclusion

The ICC and the OTP must welcome the entrance of ordinary civilians as players in human rights investigations. It is in the OTP’s best interest to prioritize the successful use of open-source information in order to ensure investigations may proceed past evidentiary thresholds.33 The Court should remain hesitant to over-rely on any one piece of open-source content, while, at the same time, adopting a verification scheme with the assistance of third-party organizations that prioritizes and safeguards the rights of all parties subject to the tribunal’s investigative arm. A failure to do so would be a failure of the Court’s critical obligation to evolve to the modern age.

Endnotes — (click the footnote reference number, or ↩ symbol, to return to location in text).

  1. 1.

    The Prosecutor v. Mahmoud Mustafa Busyf Al-Werfalli, ICC-01/11-01/17, Warrant of Arrest (PTC I, Aug. 15, 2017), available online; Alexa Koenig, Open Source Evidence and Human Rights Cases: A Modern Social History, in Digital Witness: Using Open Source Information for Human Rights Investigation, Documentation, and Accountability 32, 40 (Sam Dubberley, Alexa Koenig & Daragh Murray, eds., Feb. 19, 2020), paywall, archived. ↩

  2. 2.

    See Alexa Koenig, “Half the Truth is often a Great Lie”: Deep Fakes, Open Source Information, and International Criminal Law, 113 Am. J. Int’l L. Unbound 250 (Aug. 19, 2019), available online, doi. ↩

  3. 3.

    Id. ↩

  4. 4.

    See, e.g., Danielle Keats Citron, Professor of Law, House Intelligence Committee Hearing on the National Security Challenge of Artificial Intelligence, Manipulated Media, and “Deep Fakes”, 2 (Jun. 13, 2019) [hereinafter Citron Testimony], available online, archived. ↩

  5. 5.

    Id. ↩

  6. 6.

    Rebecca J. Hamilton, User-Generated Evidence, 57 Colum. J. Transnat’l L. 1 (Nov. 7, 2018), available online. ↩

  7. 7.

    Koenig, supra note 2, at 252. ↩

  8. 8.

    Id. ↩

  9. 9.

    Citron Testimony, supra note 4, at 2. ↩

  10. 10.

    Rachel Metz, The Number of Deepfake Videos Online is Spiking. Most are Porn, CNN, Oct. 7, 2019, available online. ↩

  11. 11.

    Edelman Trust Barometer 2018—UK Findings, Edelman (Jan. 22, 2018), available online. ↩

  12. 12.

    Hamilton, supra note 6, at 18. ↩

  13. 13.

    Paul Holmes, Reuters Toughens Rules after Altered Photo Affair, Reuters, Jan. 18, 2007, available online. ↩

  14. 14.

    Mike Nizza & Patrick J. Lyons, In an Iranian Image, a Missile Too Many, N.Y. Times, Jul. 10, 2008, available online. ↩

  15. 15.

    Conor Fortune, Digitally Dissecting Atrocities: Amnesty International’s Open Source Investigations, Amnesty Int’l (Sep. 26, 2018), available online. ↩

  16. 16.

    Paola Verhaert, Amnesty International’s Digital Verification Corps: New Networks and Methods for Human Rights Research, The Engine Room (Jun. 19, 2017), available online. ↩

  17. 17.

    Fortune, supra note 15. ↩

  18. 18.

    Verhaert, supra note 16. ↩

  19. 19.

    Id. ↩

  20. 20.

    Id. ↩

  21. 21.

    Koenig, supra note 1, at 44. ↩

  22. 22.

    Id. ↩

  23. 23.

    Id. at 32. ↩

  24. 24.

    About, Bellingcat, available online (last visited May 25, 2020). ↩

  25. 25.

    Id. ↩

  26. 26.

    Koenig, supra note 1, at 44. ↩

  27. 27.

    Digital Forensic Research Lab, Atlantic Council, available online (last visited May 25, 2020). ↩

  28. 28.

    Id. ↩

  29. 29.

    See Nancy Amoury Combs, Investigative Delegations: Predictable Predicaments, 113 Am. J. Int’l L. Unbound 267, 270 (Aug. 19, 2019), available online, doi

    (citing Hamilton, supra note 6, 39–42). ↩

  30. 30.

    Id. ↩

  31. 31.

    Bobby Chesney & Danielle Keats Citron, Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security, 107 Cal. L. Rev. 1753, 1758 (2019), available online, doi. ↩

  32. 32.

    Verhaert, supra note 16. ↩

  33. 33.

    Hamilton, supra note 6, at 12.

    ↩
Collapse