Dynamic program slicing is widely recognized as a powerful aid for e.g. Program comprehension during debugging. However, its widespread use has been impeded in part by scalability issues that occur when constructing the dynamic dependence graph necessary to compute dynamic slices. A few seconds of execution time on a modern CPU can easily yield dynamic dependence graphs on the order of tens of gigabytes in size. Existing methods either produce imprecise slices, incur large time overheads during slice computation, or run out of memory for long program executions. By carefully designing our method to take advantage of locality, we are able to efficiently use secondary storage for dynamic dependence graphs, thus allowing our method to scale to long program executions. Our prototype implementation runs directly on x86 executables, eliminating problems with e.g. Binary-only libraries. We show in our experiments that graphs can be constructed for program runs with billions of executed instructions, at slowdowns ranging from 62x to 173x. Our optimized format also allows graphs to be traversed at speeds of several million dependence edges per second.

Program trace alignment is the process of establishing a correspondence between dynamic instruction instances in executions of two semantically similar but syntactically different programs. In this paper we present what is, to the best of our knowledge, the first method capable of aligning realistically long execution traces of real programs. To maximize generality, our method works entirely on the machine code level, i.e. it does not require access to source code. Moreover, the method is based entirely on dynamic analysis, which avoids the many challenges associated with static analysis of binary code, and which additionally makes our approach inherently resilient to e.g. static code obfuscation. Therefore, we believe that our trace alignment method could prove to be a useful aid in many program analysis tasks, such as debugging, reverse-engineering, investigating plagiarism, and malware analysis. We empirically evaluate our method on 11 popular Linux programs, and show that it is capable of producing meaningful alignments in the presence of various code transformations such as optimization or obfuscation, and that it easily scales to traces with tens of millions of instructions.

Mutation-based fuzzing is a popular and widely employed black-box testing technique for finding security and robustness bugs in software. It owes much of its success to its simplicity; a well-formed seed input is mutated, e.g. through random bit-flipping, to produce test inputs. While reducing the need for human effort, and enabling security testing even of closed-source programs with undocumented input formats, the simplicity of mutation-based fuzzing comes at the cost of poor code coverage. Often millions of iterations are needed, and the results are highly dependent on configuration parameters and the choice of seed inputs. In this paper we propose a novel method for automated generation of high-coverage test cases for robustness testing. Our method is based on the observation that, even for closed-source programs with proprietary input formats, an implementation that can generate well-formed inputs to the program is typically available. By systematically mutating the program code of such generating programs, we leverage information about the input format encoded in the generating program to produce high-coverage test inputs, capable of reaching deep states in the program under test. Our method works entirely at the machine-code level, enabling use-cases similar to traditional black-box fuzzing. We have implemented the method in our tool MutaGen, and evaluated it on 7 popular Linux programs. We found that, for most programs, our method improves code coverage by one order of magnitude or more, compared to two well-known mutation-based fuzzers. We also found a total of 8 unique bugs.

Greybox fuzzing has recently emerged as a scalable and practical approach to finding security bugs in software. For example, AFL — the current state-of-the-art greybox fuzzer — has found hundreds of vulnerabilities in popular software since its release in 2013. The combination of lightweight coverage instrumentation and a simple evolutionary algorithm allows AFL to quickly generate inputs that exercise new code. AFL also obviates the need to manually set ad-hoc fuzzing ratios, which has been a major limitation of classical black-box fuzzers. Instead, AFL's first fuzzing pass exhaustively applies a set of mutations to every byte of a program input. While this approach allows for more thorough exploration of the input space, and therefore improves the chances of finding complex bugs, it also drastically slows down the fuzzing progress for "heavyweight" programs, or programs that take large inputs. This makes AFL less suitable for fuzzing input formats with large size overhead, such as various document formats. In this paper, we propose focused fuzzing as a practical trade-off between thoroughness and speed, for fuzzers that employ input mutation. We extend the notion of code coverage to individual bytes of input, and show how forward dynamic slicing can be used to efficiently determine the set of program instructions that are affected by a particular input byte. This information can then be used to restrict expensive mutations to a small subset of input bytes. We implement focused fuzzing on top of AFL, and evaluate it on four "real-life" Linux programs. Our evaluation shows that focused fuzzing noticeably improves bug discovery, compared to vanilla AFL.