This routing protocol calculates the trustworthiness of IP cores in network-on-chip (NoC) systems and makes routing decisions to avoid malicious cores that corrupt transmissions and hamper performance. Many system-on-chip (SoC) designs communicate via a network connecting various system cores, many of which are manufactured by multiple third-party vendors to lower cost and production time. Some untrusted vendors produce malicious cores that tamper with packet transmissions and cause multiple retransmissions and stall cycles. Efforts to limit malicious corruption of packets rely on authentication schemes, but these schemes introduce unacceptable overhead on resource constrained network-on-chip systems.
Researchers at the University of Florida have developed a trust-aware routing protocol that effectively avoids malicious IPs on a microchip. By communicating only through trusted cores, a network-on-chip system avoids transmission corruption and delay, uses less energy, and shortens encryption time.
Routing protocol to avoid malicious IPs in network-on-chip based system-on-chips
The routing protocol calculates trust relationships on a spectrum between pairs of nodes, which represent potentially-malicious IP cores in an SoC. The protocol expands and updates the nodal relationships continuously to determine the trust value of nodal paths in order to send transmissions only from secure zone to secure zone. If a malicious node corrupts a transmission, the retransmission reduces that node’s trust value and an algorithm determines an alternative path. SoCs suffer significant packet transfer delays in the presence of malicious IPs. Data from both real and synthetic benchmarks demonstrate that this protocol reduces packet transmission delays and energy consumption nearly to the level of a perfectly secure network-on-chip.
