Assessment of VLSI resources requirement for a sliced trusted platform module

Recent increases in cybercrime suggest questions such as: How can one trust a secure system? How can one protect private information from being stolen and maintain security? Trust in any system requires a foundation or root of trust. A root of trust is necessary to establish confidence that a machine is clean and that a software execution environment is secure. A root of trust can be implemented using the Trusted Platform Module (TPM), which is promising for enhancing security of general-purpose computing systems. In cloud computing, one of the proposed approaches is to use homomorphic encryption to create k program slices to be executed on k different cloud nodes. The TPM at the cloud node can then also be distributed or sliced along the lines presented in this thesis. In this work, we propose to increase TPM efficiency by distributing the TPM into multiple shares using Residue Number Systems (RNS). We then perform an evaluation of the silicon area, and execution time required for a sliced-TPM implementation and compares it to a single TPM. We characterize the execution time required by each TPM command using measurements obtained on ModelSim simulator. Finally, we show that the proposed scheme improves TPM efficiency and that execution time of TPM commands was noticeably improved. In the case of 4 shares the required execution time of the TPM commands that involving RSA operation in each slice was decreased by 93%, and the area of each slice was decreased by 2.93% while the total area was increased by 74%. In the case of 10 shares the required execution time of the TPM commands that involving RSA operations in each slice was decreased by 99%, and the area of each slice was decreased by 3.3% while the total area was increased by 85%