NIRVANA: Non-Invasive Real-Time VulnerAbility ANAlysis for RISC-V Processor

Embedded systems are increasingly susceptible to attack from malicious software, posing a significant threat to critical infrastructure and data. Various shreds of evidence reveal the unknown nature of attacks. In this manuscript, we propose a novel abnormal behaviour monitoring and detection system by designing a self-supervised hardware-based Self-Organizing Map (SOM) algorithm which continuously monitors the execution status of an embedded program and the behaviour of the entire platform as a whole. Our design boasts low resource utilization, high speed, effectiveness, and broad compatibility, making it suitable for real-time detection of malicious behaviour in resource-constrained embedded systems. Experimental trials were conducted on the Piccolo RISC-V processor being prototyped on an FPGA, which achieved an impressive 96% accuracy in detecting malicious programs, at the cost of a marginal 10% increase in resource consumption in comparison to its vanilla counterpart.