ETHPLOIT: From Fuzzing To Efficient Exploit Generation .
E TH P LOIT: From Fuzzing to Efficient ExploitGeneration against Smart ContractsQingzhao Zhang ‡ , Yizhuo Wang ‡ , Juanru Li , Siqi Ma† () ShanghaiJiao Tong University, China{fszqz001, mr.wang-yz}@sjtu.edu.cn, roman@sjtusec.com† Data 61, CSIRO, Australiasiqi.ma@csiro.auAbstract—Smart contracts, programs running on blockchainsystems, leverage diverse decentralized applications (DApps).Unfortunately, well-known smart contract platforms, Ethereumfor example, face serious security problems. Exploits to contractsmay cause enormous financial losses, which emphasize theimportance of smart contract testing. However, current exploitgeneration tools have difficulty to solve hard constraints inexecution paths and cannot simulate the blockchain behaviorsvery well. These problems cause a loss of coverage and accuracyof exploit generation.To overcome the problems, we design and implement E TH P LOIT, a smart contract exploit generator based on fuzzing.E TH P LOIT adopts static taint analysis to generate exploittargeted transaction sequences, a dynamic seed strategy to passhard constraints and an instrumented Ethereum Virtual Machineto simulate blockchain behaviors. We evaluate E TH P LOIT on45,308 smart contracts and discovered 554 exploitable contracts.E TH P LOIT automatically generated 644 exploits without anyfalse positive and 306 of them cannot be generated by previousexploit generation tools.Index Terms—smart contract, fuzzing, exploitationI. I NTRODUCTIONBlockchain-based cryptocurrency systems gain great popularity in the past several years, standing at an overall market capitalization of 170 billion dollars in April 2019 [1].Ethereum [2], for example, is the second-largest blockchainsystem and cryptocurrency after Bitcoin [3] by overall market value [1]. Ethereum develops Bitcoin’s scripts, a pieceof stack-based code performing some simple checks beforecurrency transfer, to a Turing-complete on-chain programminglanguage called smart contract. As a result, more than cryptocurrency, Ethereum serves as a platform for decentralizedapplications (i.e., DApps) based on smart contracts, suchas tokens, gambling, auction, etc. The website State of theDApps [4] records over 2,200 active DApps, over 85% ofwhich come from Ethereum. In fact, Ethereum had hosted overone million smart contracts by the end of 2018 [5].As the smart contract develops, security vulnerabilities ofEthereum smart contracts become a serious problem. TheTuring-complete language makes smart contracts more errorprone than Bitcoin’s scripts and several vulnerabilities havebeen discovered [6]. Because of the nature of cryptocurrency,smart contracts usually involve flows of ETH, the virtual‡ These authors equally contribute to the paper.currency in Ethereum which is also valuable in the real world.Therefore, when attackers make use of vulnerabilities, thecurrency is illegally manipulated. In fact, real-world attackssuch as DAO [7] and Parity Multisig Wallet [8] have causeda tremendous crisis for Ethereum. What is worse, smartcontracts are unmodifiable after on-chain deployment and thedamage of attacks is almost irretrievable.One major research goal for smart contracts vulnerabilityis how to fulfill an accurate vulnerability detection. Existingdetection techniques [9]–[12] often suffer from both falsenegative and false positive. To verify a detected vulnerability,a typical method is to generate an exploit to test whetherthe vulnerability can be actually triggered. Current exploitgeneration for smart contracts generally applies symbolicexecution method [13], [14]. Teether [13], for instance, locatespotential dangerous operations and solves an execution pathtriggering the operation using SMT solvers. However, such aworkflow of symbolic execution faces two challenges. The firstone is Unsolvable Constraints, which indicates the executionof sensitive operations (e.g., currency transfer, self-destruction)is restricted by conditions that are difficult for existing tools topass. Unsolvable Constraints is a major obstacle for symbolicexecution. A prominent case is the validity check of hashvalue before currency transfer. For instance, teether tries tojump over a hash instruction but cannot present value relationbetween hash pre-image and hash value, which does nothelp to pass cryptographic checks in many scenarios. Ourobservation on 49,522 contracts collected from etherscan [5]demonstrates that these complicated conditions are commonas 20% (9,694) contracts contain cryptographic functions. Inthis situation, the corresponding exploit is not able to begenerated. Another challenge for current techniques is how tohandle Blockchain Effects. Blockchain properties (e.g., timestamp and block number) are variables used in the contracts,which represent information about objects (e.g., block) in theblockchain system. Current exploit generation tools regardblockchain properties as normal global variables. However,these properties have their meaning in the blockchain systemand their value has a specific range. If not handled properly,it is also infeasible to generate an exploit successfully.To respond to the above challenges, in this paper, weutilize fuzzing to generate exploits. Compared with symbolicexecution, fuzzing does not need to mathematically solve
the exploit generation technique against smart contracts and developers could leverage our tool to build more secure code. In summary, we make the following contributions: 1)We summarized two major challenges, Unsolvable Con-straints and Blockchain Effects, in smart contract exploit
Fuzzing for Software Security Testing and Quality Assurance Media whore. Overview The fuzzing setup Fuzzing PDF's, Preview and Adobe Acrobat Reader Fuzzing PPT's, OpenOffice and MS PowerPoint Fuzzing "truths" revealed. About this talk Most fuzzing talks take one of two forms
Fuzzing JavaScript Engines with Aspect-preserving Mutation Soyeon Park Wen Xu Insu Yun Daehee Jang Taesoo Kim Georgia Institute of Technology {spark720, wen.xu, insu, daehee87, taesoo} @ gatech.edu Abstract—Fuzzing is a practical, widely-deployed technique to find bugs in complex, real-world programs like JavaScript engines.
in-depth view on fuzzing outlining the state-of-the-art advance-ment in this area since its original introduction. RQ2, which is discussed in Section VI, is proposed to give an insight into the scope of fuzzing and its applicability to different domains. Fi-nally, based on the answer to the previous questions, we expect
fuzzing to big data analytics directly for three reasons: (1) the long latency of DISC systems prohibits the applicability of fuzzing: naïve fuzzing would spend 98% of the time in setting up a test environ-ment; (2) conventional branch coverage is unlikely to scale to DISC applications because most binary code comes from the framework
Hartmut Pohl: Automated Testing with Commercial Fuzzing Tools 4 After the interfaces have been successfully identified, input data can be generated using a Fuzzer; these data can then be
Lecture 4: Dynamic Analysis and Fuzzing Lecturer: Suman Jana Scribe: Jonas Guan Feb 21, 2019 Presentation Logistics Starting from the class on March 7th, students will begin to indiv
with 8.35%-36.84% more paths within 24 hours, and has exposed 9 previously unknown vulnerabilities. Index Terms—Fuzzing, ICS Protocol, Vulnerability Detection I. INTRODUCTION Industrial Control System (ICS) refers to a system combining hardware and software with netwo
Alex Rider was woken by the first chime. His eyes flickered open, but for a moment he stayed completely still in his bed, lying on his back with his head resting on the pillow. He heard a bedroom door open and a creak of wood as somebody went downstairs. The bell rang a second time, and he looked at the alarm clock glowing beside him. There was a rattle as someone slid the security chain off .
