A probing approach for hardware trojan localization in noc-based manycores

• Ano de defesa: 2025
• Autor(a) principal: Rodrigues, Gustavo Comaru
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul; Escola Politécnica; Brasil; PUCRS; Programa de Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: NoC-Based Manycores; Security; Hardware Trojan (HT); HT Localization; CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO
• Link de acesso: https://tede2.pucrs.br/tede2/handle/tede/11635

As the adoption and complexity of manycore systems increase, ensuring data protection has become a critical design requirement. Additionally, the widespread use of third-party intellectual property cores (3PIPs) to meet time-to-market constraints and reduce design costs, raises the risk of malicious hardware insertion through Hardware Trojans (HTs), thereby increasing the vulnerability of manycore platforms. The Network on-Chip (NoC), due to its central role in the architecture, becomes an attractive target for HT insertion, as it provides access to all other system components. An HT infecting the NoC can enable various attacks, such as denial-of-service (DoS) and performance degradation. When such attacks are detected, the system must deploy countermeasures to halt the attack and protect running applications. However, not knowing the HT´s location reduces the effectiveness of countermeasures. Although the literature offers techniques for identifying the source of attacks in NoCs, these typically require additional security features integrated into NoC hardware, rendering them unsuitable for non-secure 3PIP-based NoCs. This dissertation aims to develop a non-invasive method for localizing HT-infected links in the NoC to address the limitation of adding hardware to non-secure modules. This work introduces a three-phase security framework executing the following actions: (1) monitors inter-task communication to detect HT attacks; (2) employs a localization algorithm to identify the infected links within the NoC; and (3) applies countermeasures to neutralize or mitigate the effects of the attack. The HT localization algorithm uses a technique called path probing, which transmits probe packets along specific NoC paths to evaluate the link integrity. The algorithm selectively sends probes and analyzes their outcomes, refining the search with each result until the HT is accurately localized. The method is implemented in software, allowing HT localization without modifying the NoC hardware. To validate the proposed approach, we conducted a series of attack campaigns in which HTs attacked the manycore using different activation patterns. The results demonstrate that the security framework successfully identified the location of the HTs while incurring minimal impact on system performance.