Microsoft Technology Licensing, LLC

Ramanathan N Venkatapathy

Jayaraman Thiagarajan

Sameet Harishanker Agarwal

Manuvir Das

Manuel Alfred Fahndrich

F Soner Terek

Hannes Helmut Ruescher

William Hunter Hudson

Dong Wei

Spencer Wong Low

Daniel W Weise

William Hiroaki Shihara

Methods and systems for analyzing a computer program use static and interprocedural analysis techniques and engines. A data processing operation, such as a function, is automatically identified within the computer program. It is determined whether the function represents a potential source for entry of untrusted data into the computer program. A course of the untrusted data is modeled through the identified function to produce a validation result, such as a call stack. Based on an attribute of the untrusted data (for example, whether the untrusted data is an unbounded integer or a string), it is determined whether the validation result identifies a security vulnerability of the computer program. A security vulnerability may exist, for example, when the modeled course of an unbounded integer through the function produces a buffer overrun in a call stack. The validation result is provided, via an API, software development tool, or user interface, for example.

Detection of security vulnerabilities in computer programs

Various techniques can be used to detect programming defects relating to the use of integers. A data structure can be created to represent ordering relationships in software instructions. Such ordering relationships can represent common unsound programmer assumptions. After annotating the data structure, unvalidated ordering relationships can be identified. Validation can use both explicit and implicit techniques. Filtering can be used to focus on only significant integer expressions, such as those used in buffer operations. For example, buffer accesses that attempt out-of-bounds buffer accesses due to integer overflow can be detected.

Defect detection for integers

Techniques and tools for implementing a source code annotation language are described. In one aspect, keywords are added to a function interface to define a contract for the function independent of function call context. In another aspect, annotations are inserted at global variables, formal parameters, return values, or user-defined types. The annotations include, for example, properties and qualifiers. A property can indicate, for example, a characteristic of a buffer. In another aspect, an annotation indicates that a value has usability properties sufficient to allow a function to rely on the value, where the usability properties depend on value type.

Source code annotation language

A preprocessor-based instrumentation technique wherein source code is provided to a compiler that exposes data structures such as symbols, types, expressions, etc. during the construction of the syntax tree in order to analyze where possible defects might make it desirable to instrument the code. Once a possible defect is detected in a segment of code, information is stored in a PI file detailing what line or lines of code to add, remove, or alter so that the code is instrumented to monitor the code segments at runtime. The PI file and source code are then provided to a preprocessor that uses the information in the PI file to insert instrumentation into the source code prior to the construction of a syntax tree. The now instrumented source code is used to build an instrumented syntax tree, which is then compiled so as to produce executable object code.

Preprocessor-based source code instrumentation

Growing community of inventors

Redmond, WA, United States of America

Ramanathan N Venkatapathy