Peter Lafosse (Owner and Co-Founder of Vector 35 Inc.)

Binary analysis serves as a foundational technique for a wide array of cybersecurity tasks, including vulnerability identification and malware analysis. While these methods have evolved to become incredibly powerful, they are inherently bounded by the limitations of what can be inferred solely from the binary data within a file. This talk aims to provide an in-depth exploration of both the capabilities and the constraints of binary analysis, dissecting the fundamental goals that drive its usage and to explore potential solutions to these constraints. We will investigate the common objectives of binary analysis, such as code understanding, bug hunting, and threat analysis, and evaluate how these goals often remain unmet when confined to mere binary introspection. This talk will argue that while binary analysis is indispensable, it cannot be the sole methodology employed for a comprehensive solution. The presentation will advocate for the integration of external data sources, contextual information, runtime behavior, and most importantly machine learning and large language models as essential components for enriching the output of binary analysis tools. By fusing binary data with external inputs, we can transcend the inherent limitations and offer a more nuanced, accurate, and actionable analysis for our users.

Bio: Peter LaFosse, is an industry veteran. He started his journey 18 years ago working at SI Government Solutions (later acquired by Raytheon) finding and exploiting software vulnerabilities, writing tools to the same effect and running teams as well. He is a recipient of a coveted DEFCON Black Badge for Capture the Flag where he was the offensive team leader. Having served his time as a defense contractor and being dissatisfied with the tools available for software reverse engineering he co-founded Vector 35 with his business partners with the aim of building the next generation of decompiler. Eight years later Binary Ninja stands as one of the most highly regarded products in the industry.

View More Papers

Enhancing Symbolic Execution by Machine Learning Based Solver Selection

Sheng-Han Wen (National Taiwan University), Wei-Loon Mow (National Taiwan University), Wei-Ning Chen (National Taiwan University), Chien-Yuan Wang (National Taiwan University), Hsu-Chun Hsiao (National Taiwan University)

Read More

GTrans: Graph Transformer-Based Obfuscation-resilient Binary Code Similarity Detection

Yun Zhang (Hunan University), Yuling Liu (Hunan University), Ge Cheng (Xiangtan University), Bo Ou (Hunan University)

Read More

A Unified Symbolic Analysis of WireGuard

Pascal Lafourcade (Universite Clermont Auvergne), Dhekra Mahmoud (Universite Clermont Auvergne), Sylvain Ruhault (Agence Nationale de la Sécurité des Systèmes d'Information)

Read More

GraphGuard: Detecting and Counteracting Training Data Misuse in Graph...

Bang Wu (CSIRO's Data61/Monash University), He Zhang (Monash University), Xiangwen Yang (Monash University), Shuo Wang (CSIRO's Data61/Shanghai Jiao Tong University), Minhui Xue (CSIRO's Data61), Shirui Pan (Griffith University), Xingliang Yuan (Monash University)

Read More