ChainGuard - Attack-Chain Security

ChainGuard is an AI-assisted smart-contract security investigation platform designed to address a key limitation of traditional vulnerability scanners: they often identify individual issues without explaining how multiple weaknesses across a repository could combine into a realistic attack path. For example, a scanner may report an access-control weakness, an unsafe external call, and an oracle dependency as separate findings. ChainGuard goes beyond isolated findings by analyzing relationships between contracts, functions, state dependencies, external calls, and security findings to reconstruct potential multi-step attack chains. The core workflow is: **SCAN → CONNECT → INVESTIGATE → VERIFY → BREAK** Users can provide a Solidity repository through a GitHub URL or ZIP upload. ChainGuard maps the repository, discovers contracts and functions, performs static analysis using Slither, extracts relationships such as call dependencies and external calls, and builds structured security context. An AI reasoning layer then investigates potential relationships using this evidence rather than relying only on raw source code. Potential findings are validated against available source and static-analysis evidence before being presented. Each potential attack chain can contain an entry point, preconditions, ordered attack steps, affected functions, source locations, supporting evidence, assumptions, limitations, confidence, and potential break points. The centerpiece is an interactive attack graph that allows users to move from an entry point through individual attack steps toward potential impact. Selecting a graph node connects the investigation to the relevant evidence and source location, allowing a security engineer to understand why that step exists and what evidence supports it. ChainGuard also provides a defender-oriented view. Instead of stopping at “a vulnerability was found,” it asks where the chain can be broken—for example, by strengthening access control, validating oracle inputs, enforcing sanity bounds, or protecting sensitive withdrawal flows. The primary target users are smart-contract developers, security researchers, auditors, and security engineers who need repository-level visibility rather than a flat list of detector results. What makes ChainGuard distinctive is the combination of **detection, connection, evidence, investigation, and break-point analysis**. Slither provides important static-analysis input, but ChainGuard treats those findings as the starting point for an investigation rather than the final product. ChainGuard deliberately presents results as potential or evidence-supported findings and preserves uncertainty where runtime exploitability has not been confirmed. It is designed to assist security investigation rather than claim to replace a complete human audit.