Patent Application
20090064118
The invention relates generally to software optimization and more particularly, to simplification of executable software programs.
Malicious software, such as viruses, worms, Trojan horse programs, spyware, and other malware, may use software obfuscation techniques to hide malicious behavior in order to make analysis and removal more difficult. Software obfuscation increases the amount of time it takes for identifying, understanding malware algorithms, which may delay the time before a fix becomes available. Software obfuscation used by malware may include unnecessary complications in instruction sequences, such a set of instructions that is effectively useless or includes an unnecessarily high number of steps, overly complex control flow, such as unnecessary jumps or opaque predicates, unnecessary use of the stack or registers, attempts to confuse debuggers regarding which bytes represent data or instructions, and other methods intended to confuse and delay a reverse engineer and/or reverse engineering tools. These techniques makes algorithms difficult to understand. Unfortunately, manual obfuscation removal is a tedious and error-prone process.
Identifying a section of target software, which matches trigger criteria, emulating the section to identify its functionality, and substituting a simpler set of data and/or instructions having equivalent functionality, allows for automated software deobfuscation. Deobfuscation may be recursive or iterated multiple times, with a first pass performing some simplification, and subsequent passes simplifying the already-simplified software even further.
Some embodiments of methods of deobfuscating software embodied on a computer readable medium comprise identifying at least one section of target software matching trigger criteria, emulating at least a portion of the identified section to determine a first function, and generating deobfuscated software by substituting a simplified section for the identified section, wherein the simplified section has a second function equivalent to the first function. A function may be a repeatable, measurable effect on computer memory. Some embodiments further comprise reading the target software from a computer readable medium and/or writing the deobfuscated software to a computer readable medium. The simplified section of software may contain one or more no operation (NOP) instructions, which creates slack space and, in some embodiments, the simplified section is the same length in bytes as the replaced identified section. Emulating the identified section of target software may comprise simulating the effect of the identified section on a memory location and/or control flow. The memory location may be a program stack, a register, a cache location, or general random access memory (RAM). Control flow may be analyzed by examining the effect of jump instructions on the execution pointer and other memory locations.
In some embodiments some or all of the target software, deobfuscated software, identified section and simplified section are represented with assembly language instructions. Identifying a section of the target software for emulation and/or simplification may involve pattern matching and/or behavior analysis of the software. A deobfuscator may use a predefined set of modes, wherein different modes use different rule sets for generating the deobfuscated software. Some modes may be more aggressive than other modes, and make more assumptions regarding the function of the software. Some embodiments insert jump instructions to bypass long sections of NOPs. An embodiment of a deobfuscation system comprises a specially-configured hardware device, such as a field programmable gate array (FPGA) or application specific integrated circuit (ASIC). A tangible, useful and concrete hardware device embodiment comprising a processor and memory takes in target software as a data stream input and generate deobfuscated software as a data stream output.
The foregoing has outlined the features and technical advantages of the invention in order that the description that follows may be better understood. Additional features and advantages of the invention will be described hereinafter. It should be appreciated by those skilled in the art that the conception and specific embodiments disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes of the invention. It should also be realized by those skilled in the art that such equivalent constructions do not depart from the spirit and scope of the invention as set forth in the claims. The novel features which are believed to be characteristic of the invention, both as to its organization and method of operation, together with further objects and advantages will be better understood from the following description when considered in connection with the accompanying figures. It is to be expressly understood, however, that each of the figures is provided for the purpose of illustration and description only and is not intended as a definition of the limits of the invention.
For a more complete understanding of the present invention, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:
A comparison of
Some of the novel aspects of automated software deobfuscation, performed in accordance with an embodiment of the invention, are sections of the target software are subject to emulation. That is, rather than the software section being sent to execute on a central processing unit (CPU) or other processor, the effects of the software section on a virtualized processor and virtualized memory are determined. Based on the determined function, a simplified section of software, having the same function, may be substituted. In some embodiments, the substitute simplified section uses a same number of bytes as the replaced section. If the simplified section uses operable instructions requiring fewer bytes, the difference is made up with no operation (NOP) instructions.
For example, a section of the target software may be analyzed and determined to have no ultimate or lasting effect on any memory locations, such as a program stack, a register, or memory that has been allocated to the process. In such a situation, the replacement simplified section would comprise a set of NOPs that is as long as the replaced section of useless instructions. In some embodiments, a jump is inserted to skip a long string of NOP instructions. In some embodiments, a long string of NOP instructions may be deleted, and other jumps recalculated to ensure that the proper destination point is reached when jumping to instructions after the removed set of NOPs.
As another example, a set of multiple jump commands, wherein at least one is a conditional jump, may be analyzed and shown to result in the same net effect, whether a conditional jump is taken or not. One implementation of this type of obfuscation is for one jump to point to a first memory location that contains a NOP, and a second jump to point to a second memory following the first memory location. If the first jump is taken the result would be that a processor receives NOP instructions until the execution pointer points to the second memory location, making the jumps to different memory locations have no different effects. In this situation, any conditional tests leading to a conditional jump are unnecessary and may be replaced with NOPs, and all jumps may then be replaced with a single jump. As another example, values may be pushed onto the program stack and then removed, using PUSH and POP instructions, such that the net effect on the program stack memory is nulled out. The PUSH and POP instructions, along with the data, may then be replaced with NOPs. These NOPs create slack space, which may be used in the event that a simplified set of instructions actually requires more bytes of memory. As yet another example, PUSH and POP instructions may be replaced with a single MOV instruction, under certain conditions.
In decision block 308, if a match is determined between the tested instruction set in the target software and trigger criteria, the instruction set is identified as obfuscated software. If, in decision block 308, a match is not detected, method 300 moves to decision block 316 to determine whether another section of the target software needs to be tested against the trigger criteria.
Obfuscation can take many forms, including the incorporation of useless and confusing instructions, mangled jumps, unnecessary data cross-references, and other techniques, such as anti-disassembly techniques that are designed to prevent the generation of an assembly language representation of the software. For example the identified section may PUSH a value onto the stack, POP it into a register, such as EAX, perform a math operation on the contents of EAX, such as an XOR, and then JMP to the contents of EAX. The function of this identified section is merely to jump to a calculated address, and could be replaced with a simply JMP instruction with the same calculated value, followed by NOPs to replace the excess number of bytes used by the original set of instructions.
Another identified section could include a series of NOPs with a JZ and a JMP instruction to various locations within the string of NOPs. No matter which jump is followed, the end result is effectively the same. Thus, the JZ and JMP are useless instructions. Some obfuscation will include alternate conditional jumps, JZ and JNZ, to the same location, making the condition check useless. Such an identified section has a function of jumping to a certain location, no matter what the value was used in the condition check. If a register is used for math operations, and the result is not used in a meaningful manner, the math instructions may be useless and are candidates for replacement with NOPs. It should be understood, though, that deobfuscation may require multiple iterations, if for example, one pass through the target software creates simplifications that enables further simplification during a subsequent pass.
Some examples of anti-disassembly include jumping into the middle of an instruction or a data value, which then alters which bytes are identified by a disassembler as instructions versus which bytes are interpreted as data. Such a change could drastically alter a control flow graph, and allow progress in understanding the algorithm.
In block 310, the identified section, as determined in decision block 308, is emulated to determine its function, for example to identify its effect on a memory location and/or control flow. The emulation tracks register math to determine the end result of calculations, so that, in some situations, the end result may be used in place of the instructions used to calculate it. The emulation also determines jump locations and function calls, and tracks register and stack operations. In block 312, simplified code is generated having the same function as determined during emulation. The simplified code has the same number of bytes as the identified section, using NOPs to create slack space when the simplified section uses fewer bytes for instructions. The slack space may be used in subsequent iterations of the deobfuscation, in the event that additional bytes are needed to substitute a simplified instruction. In block 314, the simplified section is injected over the identified section with a binary injector. In some embodiments, the binary injector may comprise a PERL script.
In decision block 316, method 300 determines whether more sections of the software require testing for deobfuscation. If so, method 300 returns to block 306. Otherwise, method 300 proceeds to decision block 318 to determine whether the deobfuscation process needs to be iterated again from block 304. Stopping criteria may be automated, for example by using a threshold for the number of substitutions during the previous pass, or it may be determined manually by user interaction. Method 300 may use IDA Pro for interfacing with a user and the target software, and thus present the user with a control graph and disassembly results so that the user can determine whether another iteration of deobfuscation is desired. At block 320, the slack space is bypassed by inserting jump instructions to skip long sections of NOPs.
The deobfuscation process of method 300 is controlled by a rule set, and the settings are selected in block 324. In one embodiment, the rule set has five mode settings: (1) Anti-disassembly—replace anti-disassembly with simplified code; (2) Passive—use safe assumptions about memory content changes and usage; (3) Aggressive—use more aggressive assumptions about memory content changes and usage; (4) Ultra—uses even more aggressive assumptions about memory content changes and usage; (5) Remove NOPs—optionally remove slack space. In block 322, the deobfuscated software is output, for example by writing the target software to a computer readable medium or by producing an output data stream.
Although the present invention and its advantages have been described, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims. Moreover, the scope of the present application is not intended to be limited to the particular embodiments of the process, means, methods and steps described in the specification. As one of ordinary skill in the art will readily appreciate from the disclosure of the present invention, processes, machines, manufacture, compositions of matter, means, methods, or steps, presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding embodiments described herein may be utilized according to the present invention. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.
This application claims priority from U.S. Provisional Application No. 60/968,569, entitled “SOFTWARE DEOBFUSCATION SYSTEM AND METHOD”, filed on Aug. 29, 2007, the disclosure of which is hereby incorporated in its entirety by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60968569 | Aug 2007 | US |
