The energy allocation strategy is one of the most popular techniques in fuzzing to improve code coverage and vulnerability discovery. The core intuition is that fuzzers should allocate more computational energy to the seed files that have high efficiency to trigger unique paths and crashes after mutation. Existing solutions usually define several properties, e.g., the execution speed, the file size, and the number of the triggered edges in the control flow graph, to serve as the key measurements in their allocation logics to estimate the potential of a seed. The efficiency of a property is usually assumed to be the same across different programs. However, we find that this assumption is not always valid. As a result, the state-of-the-art energy allocation solutions with static energy allocation logics are hard to achieve desirable performance on different programs.

To address the above problem, we propose a novel program-sensitive solution, named SLIME, to enable adaptive energy allocation on the seed files with various properties for each program. Specifically, SLIME first designs multiple property-aware queues, with each queue containing the seed files with a specific property. Second, to improve the return of investment, SLIME leverages a customized Upper Confidence Bound Variance-aware (UCB-V) algorithm to statistically select a property queue with the most estimated reward, i.e., finding the most new unique execution paths and crashes. Finally, SLIME mutates the seed files in the selected property queue to perform property-adaptive fuzzing on a program. We evaluate SLIME against state-of-the-art open source fuzzers AFL, MOPT, AFL++, AFL++HIER, EcoFuzz, and TortoiseFuzz on 9 real-world programs. The results demonstrate that SLIME discovers up to 3.53X more unique vulnerabilities than the baselines, and achieves better coverage performance. We will open source a prototype of SLIME to facilitate future fuzzing research.