| Summary: | Following proper guidelines and recommendations are crucial in software security, which is mostly obstructed by accidental human errors. Automatic screening tools have great potentials to reduce the gap between the theory and the practice. However, the goal of scalable automated code screening is largely hindered by the practical difficulty of reducing false positives without compromising analysis quality. To enable compile-time security checking of cryptographic vulnerabilities, I developed highly precise static analysis tools (CryptoGuard and TaintCrypt) that developers can use routinely. The main technical enabler for CryptoGuard is a set of detection algorithms that refine program slices by leveraging language-specific insights, where TaintCrypt relies on symbolic execution-based path-sensitive analysis to reduce false positives. Both CryptoGuard and TaintCrypt uncovered numerous vulnerabilities in real-world software, which proves the effectiveness. Oracle has implemented our cryptographic code screening algorithms for Java in its internal code analysis platform, Parfait, and detected numerous vulnerabilities that were previously unknown. I also designed a specification language named SpanL to easily express rules for automated code screening. SpanL enables domain experts to create domain-specific security checking. Unfortunately, tools and guidelines are not sufficient to ensure baseline security in internet-wide ecosystems. I found that the lack of proper compliance checking induced a huge gap in the payment card industry (PCI) ecosystem. I showed that none of the PCI scanners (out of 6), we tested are fully compliant with the guidelines, issuing certificates to merchants that still have major vulnerabilities. Consequently, 86% (out of 1,203) of the e-commerce websites we tested, are non-compliant. To improve the testbeds in the light of our work, the PCI Security Council shared a copy of our PCI measurement paper to the dedicated companies that host, manage, and maintain the PCI certification testbeds. === Doctor of Philosophy === Automatic screening tools have great potentials to reduce the gap between the theory and the practice of software security. However, the goal of scalable automated code screening is largely hindered by the practical difficulty of reducing false positives without compromising analysis quality. To enable compile-time security checking of cryptographic vulnerabilities, I developed highly precise static analysis tools (CryptoGuard and TaintCrypt) that developers can use routinely. Both CryptoGuard and TaintCrypt uncovered numerous vulnerabilities in real-world software, which proves the effectiveness. Oracle has implemented our cryptographic code screening algorithms for Java in its internal code analysis platform, Parfait, and detected numerous vulnerabilities that were previously unknown. I also designed a specification language named SpanL to easily express rules for automated code screening. SpanL enables domain experts to create domain-specific security checking. Unfortunately, tools and guidelines are not sufficient to ensure baseline security in internet-wide ecosystems. I found that the lack of proper compliance checking induced a huge gap in the payment card industry (PCI) ecosystem. I showed that none of the PCI scanners (out of 6), we tested are fully compliant with the guidelines, issuing certificates to merchants that still have major vulnerabilities. Consequently, 86% (out of 1,203) of the e-commerce websites we tested, are non-compliant. To improve the testbeds in the light of our work, the PCI Security Council shared a copy of our PCI measurement paper to the dedicated companies that host the PCI certification testbeds.
|
|---|
