We propose Angora, a new mutation-based fuzzer that outperforms the state-of-the-art fuzzers by a wide margin. The main goal of Angora is to increase branch coverage by solving path constraints without symbolic execution. To solve path constraints efﬁciently, we introduce several key techniques: scalable byte-level taint tracking, context-sensitive branch count, search based on gradient descent, and input length exploration.
We studied the risks of UAVs and conducted an empirical analysis of three popular DJI UAVs. We discovered a series of vulnerabilities, including insecure communication channels, misuse of cryptography, and insecure UAV activation and developer authorization.
In this work, we explored neural network based approaches (Recurrent neural network) towards the source code author deanonymization problem. With a dataset extracted from Google Code Jam, our char-level model performs competitively on normal size dataset comparing to previous state-of-art work.