Automatic algorithm discovery using reverse dataflow analysis
View Patent ↗A system and method for finding vulnerabilities and tracing an end result associated with a vulnerability to its origins in user data. A user data source containing an ordered sequence of user data items may be a data file. In one embodiment the method for identifying, in the user data source, the origins of the end result, includes associating with each user data item a location identifier identifying the location of the user data item in the sequence of user data items executing instructions with a virtual machine, associating with each result the location identifier of the data item when one argument is a user data item and a tag when more than one argument is a user data item. This process may be continued until the end result is obtained. Subsequently, the method may include stepping through instructions with the virtual machine in reverse order, to tracing the origins of the end result to each of the user data items contributing to the result.
1. A method for identifying origins, in a user data source containing an ordered sequence of user data items, of an end result calculated from the user data items, by a virtual machine executing on a computing device, the method comprising:
retrieving a plurality of user data items from the user data source;
associating with each of the plurality of user data items, a location identifier identifying a location of the user data item in the sequence of user data items;
executing a plurality of instructions with the virtual machine, each of the plurality of instructions taking one or more arguments and producing one or more results,
associating with each result:
the location identifier of the user data item when one argument is a user data item; and
a tag when more than one argument is a user data item;
until the end result is obtained;
executing instructions with the virtual machine in reverse order, taking results and producing arguments, and
tracing the origin of a result associated with a tag to each of the user data items contributing to the result.
2. The method of claim 1 , wherein the user data source is a generated data file.
3. The method of claim 1 , wherein the user data source is user keyboard input.
4. The method of claim 1 , wherein the user data source is a source of network packets.
5. The method of claim 2 , wherein the generated data file is generated by pseudorandom modification of a retrieved data file.
6. The method of claim 1 , wherein the executing of a plurality of instructions with the virtual machine comprises periodically saving the state of the virtual machine.
7. The method of claim 6 , wherein the executing instructions in reverse order comprises returning to a previously saved state of the virtual machine and executing instructions with the virtual machine starting from the previously saved state.
8. The method of claim 1 , comprising tracing the origins of the end result to each user data item contributing to the end result.
9. The method of claim 8 , wherein the tracing of the origins of the end result to each end user data item contributing to the result comprises recursively tracing the origins of each computed user data item contributing to the end result.
10. A system for identifying origins, in a user data source containing an ordered sequence of user data items, of an end result calculated, from the user data items, by a virtual machine executing on a computing device, the system comprising:
a processor; and
a memory, wherein the memory has stored thereon instructions that, when executed by the processor, cause the processor to:
retrieve a plurality of user data items from the user data source;
associate with each of the plurality of user data items, a location identifier identifying a location of the user data item in the sequence of user data items;
execute a plurality of instructions with the virtual machine, each of the plurality of instructions taking one or more arguments and producing one or more results,
associate with each result:
the location identifier of the user data item when one argument is a user data item; and
a tag when more than one argument is a user data item;
until the end result is obtained;
execute instructions with the virtual machine in reverse order, taking results and producing arguments, and
trace the origin of a result associated with a tag to each of the user data items contributing to the result.
11. The system of claim 10 , wherein the user data source is a generated data file.
12. The system of claim 11 , wherein the generated data file is generated by pseudorandom modification of a retrieved data file.
13. The system of claim 10 , wherein the executing of a plurality of instructions with the virtual machine comprises periodically saving the state of the virtual machine.
14. The system of claim 13 , wherein the executing instructions in reverse order comprises returning to a previously saved state of the virtual machine and executing instructions with the virtual machine starting from the previously saved state.
15. The system of claim 10 , wherein the instructions cause the processor to trace the origins of the end result to each user data item contributing to the end result.
16. The system of claim 15 , wherein the tracing of the origins of the end result to each end user data item contributing to the result comprises recursively tracing the origins of each computed user data item contributing to the end result.