Sara Tucci-Piergiovanni

Zaynah Dargaye, Önder Gürcan, Florent Kirchner et al.

Distributed immutable ledgers, or blockchains, allow the secure digitization of evidential transactions without relying on a trusted third-party. Evidential transactions involve the exchange of any form of physical evidence, such as money, birth certificate, visas, tickets, etc. Most of the time, evidential transactions occur in the context of complex procedures, called evidential protocols, among physical agents. The blockchain provides the mechanisms to transfer evidence, while smart contracts - programs executing within the blockchain in a decentralized and replicated fashion - allow encoding evidential protocols on top of a blockchain. As a smart contract foregoes trusted third-parties and runs on several machines anonymously, it constitutes a highly critical program that has to be secure and trusted-by-design. While most of the current smart contract languages focus on easy programmability, they do not directly address the need of guaranteeing trust and accountability, which becomes a significant issue when evidential protocols are encoded as smart contracts.

Önder Gürcan, Nataliya Yakymets, Sara Tucci-Piergiovanni et al.

Failure Mode, Effects and Criticality Analysis (FMECA) is one of the safety analysis methods recommended by most of the international standards. The classical FMECA is made in a form of a table filled in either manually or by using safety analysis tools. In both cases, the design engineers have to choose the trade-offs between safety and other development constraints. In the case of complex cyber-physical systems (CPS) with thousands of specified constraints, this may lead to severe problems and significantly impact the overall criticality of CPS. In this paper, we propose to adopt optimization techniques to automate the decision making process conducted after FMECA of CPS. We describe a multi-agent based optimization method which extends classical FMECA for offering optimal solutions in terms of criticality and development constraints of CPS.

Issam Seddik, Sami Souihi, Mohamed Tamaazousti et al.

As Large Language Models (LLMs) gain traction across critical domains, ensuring secure and trustworthy training processes has become a major concern. Backdoor attacks, where malicious actors inject hidden triggers into training data, are particularly insidious and difficult to detect. Existing post-training verification solutions like Proof-of-Learning are impractical for LLMs due to their requirement for full retraining, lack of robustness against stealthy manipulations, and inability to provide early detection during training. Early detection would significantly reduce computational costs. To address these limitations, we introduce Proof-of-Training Steps, a verification protocol that enables an independent auditor (Alice) to confirm that an LLM developer (Bob) has followed the declared training recipe, including data batches, architecture, and hyperparameters. By analyzing the sensitivity of the LLMs' language modeling head (LM-Head) to input perturbations, our method can expose subtle backdoor injections or deviations in training. Even with backdoor triggers in up to 10 percent of the training data, our protocol significantly reduces the attacker's ability to achieve a high attack success rate (ASR). Our method enables early detection of attacks at the injection step, with verification steps being 3x faster than training steps. Our results highlight the protocol's potential to enhance the accountability and security of LLM development, especially against insider threats.

Pierre-Yves Piriou, Olivier Boudeville, Gilles Deleuze et al.

In the industry, blockchains are increasingly used as the backbone of product and process traceability. Blockchain-based traceability participates in the demonstration of product and/or process compliance with existing safety standards or quality criteria. In this perspective, services and applications built on top of blockchains are business-critical applications, because an intended failure or corruption of the system can lead to an important reputation loss regarding the products or the processes involved. The development of a blockchain-based business-critical application must be then conducted carefully, requiring a thorough justification of its dependability and security. To this end, this paper encourages an engineering perspective rooted in well-understood tools and concepts borrowed from the engineering of safety-critical systems. Concretely, we use a justification framework, called CAE (Claim, Argument, Evidence), by following an approach based on assurance cases, in order to provide convincing arguments that a business-critical blockchain-based application is dependable and secure. The application of this approach is sketched with a case study based on the blockchain HYPERLEDGER FABRIC.

Zaynah Dargaye, Antonella Pozzo, Sara Tucci-Piergiovanni

The paper presents Pluralize a formal logical framework able to extend the execution of blockchain transactions to events coming from external oracles, like external time, sensor data, human-made declarations, etc. These events are by essence non-reliable, since transaction execution can be triggered by information whose veracity cannot be established by the blockchain. To overcome this problem, the language features a first-order logic and an authority algebra to allow formal reasoning and establish accountability of agents for blockchain-enabled transactions. We provide an accountability model that allows to formally prove the accountability of agents by a formal proof locally executable by each agent of the blockchain.