Line-level Semantic Structure Learning for Code Vulnerability Detection
Software code vulnerabilities refer to weaknesses in code that can be exploited, leading to severe consequences such as unauthorized information disclosure and cyber extortion. The growing magnitude of this issue is highlighted by recent statistics: in the first quarter of 2022, the US National Vulnerability Database (NVD) reported 8,051 vulnerabilities, a 25% increase from the previous year. Additionally, a study found that 81% of 2,409 analyzed codebases contained at least one known open-source vulnerability.The widespread nature and increasing number of these vulnerabilities underscore the urgent need for robust automated vulnerability detection mechanisms. Implementing such systems is crucial for enhancing software security and preventing a range of potential threats.Existing literature on vulnerability detection models can be broadly categorized into two main types: (1) traditional detection models, (2) deep learning (DL)-based models .Traditional detection models often require experts to manually develop detection rules.This approach is labor-intensive and struggle to maintaining low rates of false positives and false negatives. In contrast, deep learning (DL)-based detection methods learn vulnerability patterns from training datasets, eliminating the need for manual heuristics and enabling autonomous feature identification.They avoid manual heuristic methods and autonomously learn and identify vulnerability features.
Sat 21 JunDisplayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change
14:00 - 15:30 | Session5: Software Vulnerability and Security IINew Idea Track / Research Track at Cosmos 3A Chair(s): Chuanyi Li Nanjing University | ||
14:00 15mTalk | Devmp: A Virtual Instruction Extraction Method for Commercial Code Virtualization Obfuscators Research Track Shenqianqian Zhang Key Laboratory of Cyberspace Security, Ministry of Education, Weiyu Dong Information Engineering University, Jian Lin Information Engineering University | ||
14:15 15mTalk | Line-level Semantic Structure Learning for Code Vulnerability Detection Research Track Ziliang Wang Peking University, Ge Li Peking University, Jia Li Tsinghua University, Yihong Dong Peking University, Yingfei Xiong Peking University, Zhi Jin Peking University | ||
14:30 15mTalk | SLVHound: Static Detection of Session Lingering Vulnerabilities in Modern Java Web Applications Research Track Haining Meng SKLP, Institute of Computing Technology, CAS, China; University of Chinese Academy of Sciences, China, Jie Lu SKLP, Institute of Computing Technology, CAS, China; University of Chinese Academy of Sciences, China, Yongheng Huang Institute of Computing Technology at Chinese Academy of Sciences; University of Chinese Academy of Sciences, Lian Li Institute of Computing Technology at Chinese Academy of Sciences; University of Chinese Academy of Sciences | ||
14:45 15mTalk | Def-VAE: Identifying Adversarial Inputs with Robust Latent Representations Research Track Chengye Li Institute of Software, Chinese Academy of Sciences, Changshun Wu Université Grenoble Alpes, Rongjie Yan Institute of Software at Chinese Academy of Sciences; University of Chinese Academy of Sciences | ||
15:00 15mTalk | Fuzzing for Stateful Protocol Programs Based on Constraints between States and Message Types Research Track Kunpeng Jian Institute of Information Engineering, Chinese Academy of Sciences, Yanyan Zou Institute of Information Engineering, Chinese Academy of Sciences, Menghao Li Institute of Information Engineering, Chinese Academy of Sciences, Wei Huo Institute of Information Engineering at Chinese Academy of Sciences | ||
15:15 10mTalk | PriceSleuth: Detecting DeFi Price Manipulation Attacks in Smart Contracts Using LLM and Static Analysis New Idea Track Hao Wu Xi'an JiaoTong University, Haijun Wang Xi'an Jiaotong University, Shangwang Li Xi'an Jiaotong University, Yin Wu Xi'an Jiaotong University, Ming Fan Xi'an Jiaotong University, Yitao Zhao Yunnan Power Grid Co., Ltd, Ting Liu Xi'an Jiaotong University Pre-print |
Cosmos 3A is the first room in the Cosmos 3 wing.
When facing the main Cosmos Hall, access to the Cosmos 3 wing is on the left, close to the stairs. The area is accessed through a large door with the number “3”, which will stay open during the event.