Radina Stoykova
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Radina Stoykova has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
Katrin Franke
3 shared publications- Stefan Axelsson
Department of Computer and Systems Sciences, Stockholm University, Kista, Sweden
3 shared publications - An Thi Nguyen
Norwegian University of Science and Technology
1 shared publication - Aya Fukami
Netherlands Forensic Institute, The Hague, The Netherlands
1 shared publication - Eric Arazo
CeADAR: Ireland's Centre for AI, University College Dublin, Belfield Dublin, D04 V1W8, Ireland
1 shared publication Fergus Toolan
1 shared publication
Latest Publications
- 20 November 2025
- Published by SAGE Publications in New Journal of European Criminal Law
- Vol. 16 (4) , 381-384
Future research directions: Can European criminal law survive the AI revolution?
- 20 November 2025
- Published by SAGE Publications in New Journal of European Criminal Law
- Vol. 16 (4) , 381-384
Future research directions: Can European criminal law survive the AI revolution?
- Open Access
Emergent AI Surveillance: Overlearned Person Re-Identification and Its Mitigation in Law Enforcement Context
Generic instance search models can dramatically reduce the manual effort required to analyze vast surveillance footage during criminal investigations by retrieving specific objects of interest to law enforcement. However, our research reveals an unintended emergent capability: through overlearning, these models can single out specific individuals even when trained on datasets without human subjects. This capability raises concerns regarding identification and profiling of individuals based on their personal data, while there is currently no clear standard on how de-identification can be achieved. We evaluate two technical safeguards to curtail a model's person re-identification capacity: index exclusion and confusion loss. Our experiments demonstrate that combining these approaches can reduce person re-identification accuracy to below 2% while maintaining 82% of retrieval performance for non-person objects. However, we identify critical vulnerabilities in these mitigations, including potential circumvention using partial person images. These findings highlight urgent regulatory questions at the intersection of AI governance and data protection: How should we classify and regulate systems with emergent identification capabilities? And what technical standards should be required to prevent identification capabilities from developing in seemingly benign applications?
Open AccessEmergent AI Surveillance: Overlearned Person Re-Identification and Its Mitigation in Law Enforcement Context
Generic instance search models can dramatically reduce the manual effort required to analyze vast surveillance footage during criminal investigations by retrieving specific objects of interest to law enforcement. However, our research reveals an unintended emergent capability: through overlearning, these models can single out specific individuals even when trained on datasets without human subjects. This capability raises concerns regarding identification and profiling of individuals based on their personal data, while there is currently no clear standard on how de-identification can be achieved. We evaluate two technical safeguards to curtail a model's person re-identification capacity: index exclusion and confusion loss. Our experiments demonstrate that combining these approaches can reduce person re-identification accuracy to below 2% while maintaining 82% of retrieval performance for non-person objects. However, we identify critical vulnerabilities in these mitigations, including potential circumvention using partial person images. These findings highlight urgent regulatory questions at the intersection of AI governance and data protection: How should we classify and regulate systems with emergent identification capabilities? And what technical standards should be required to prevent identification capabilities from developing in seemingly benign applications?
- Open Access
- 5 December 2024
- Published by Elsevier in Forensic Science International: Synergy
- Vol. 9, 100563
The AI Act in a law enforcement context: The case of automatic speech recognition for transcribing investigative interviews
- 5 December 2024
- Published by Elsevier in Forensic Science International: Synergy
- Vol. 9, 100563
Open AccessThe AI Act in a law enforcement context: The case of automatic speech recognition for transcribing investigative interviews
- Top Cited PapersOpen Access
- 4 October 2024
- Published by Elsevier in Computer Law & Security Review
- Vol. 55
A New Right to Procedural Accuracy: A Governance Model for Digital Evidence in Criminal Proceedings
- 4 October 2024
- Published by Elsevier in Computer Law & Security Review
- Vol. 55
Top Cited PapersOpen AccessA New Right to Procedural Accuracy: A Governance Model for Digital Evidence in Criminal Proceedings
- Open Access
The Dilemma of Uncertainty Estimation for General Purpose AI in the EU AI Act
The AI act is the European Union-wide regulation of AI systems. It includes specific provisions for general-purpose AI models which however need to be further interpreted in terms of technical standards and state-of-art studies to ensure practical compliance solutions. This paper examines the AI act requirements for providers and deployers of general-purpose AI and further proposes uncertainty estimation as a suitable measure for legal compliance and quality assurance in training of such models. We argue that uncertainty estimation should be a required component for deploying models in the real world, and under the EU AI Act, it could fulfill several requirements for transparency, accuracy, and trustworthiness. However, generally using uncertainty estimation methods increases the amount of computation, producing a dilemma, as computation might go over the threshold ($10^{25}$ FLOPS) to classify the model as a systemic risk system which bears more regulatory burden.
Open AccessThe Dilemma of Uncertainty Estimation for General Purpose AI in the EU AI Act
The AI act is the European Union-wide regulation of AI systems. It includes specific provisions for general-purpose AI models which however need to be further interpreted in terms of technical standards and state-of-art studies to ensure practical compliance solutions. This paper examines the AI act requirements for providers and deployers of general-purpose AI and further proposes uncertainty estimation as a suitable measure for legal compliance and quality assurance in training of such models. We argue that uncertainty estimation should be a required component for deploying models in the real world, and under the EU AI Act, it could fulfill several requirements for transparency, accuracy, and trustworthiness. However, generally using uncertainty estimation methods increases the amount of computation, producing a dilemma, as computation might go over the threshold ($10^{25}$ FLOPS) to classify the model as a systemic risk system which bears more regulatory burden.
- Open Access
- 22 July 2023
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 46
Encrochat: The hacker with a warrant and fair trials?
- 22 July 2023
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 46
Open AccessEncrochat: The hacker with a warrant and fair trials?
- Open Access
- 11 May 2023
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 45
Reliability validation enabling framework (RVEF) for digital forensics in criminal investigations
- 11 May 2023
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 45
Open AccessReliability validation enabling framework (RVEF) for digital forensics in criminal investigations
- Top Cited PapersOpen Access
- 25 March 2023
- Published by Elsevier in Computer Law & Security Review
- Vol. 49
The right to a fair trial as a conceptual framework for digital evidence rules in criminal investigations
- 25 March 2023
- Published by Elsevier in Computer Law & Security Review
- Vol. 49
Top Cited PapersOpen AccessThe right to a fair trial as a conceptual framework for digital evidence rules in criminal investigations
- Open Access
- 6 August 2022
- Published by Elsevier in Computer Law & Security Review
- Vol. 46
Legal and technical questions of file system reverse engineering
- 6 August 2022
- Published by Elsevier in Computer Law & Security Review
- Vol. 46
Open AccessLegal and technical questions of file system reverse engineering
- Open Access
- 17 March 2022
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 40
Reliability assessment of digital forensic investigations in the Norwegian police
- 17 March 2022
- Published by Elsevier in Forensic Science International: Digital Investigation
- Vol. 40
Open AccessReliability assessment of digital forensic investigations in the Norwegian police