Protecting Software Code by Guards

Abstract

Abstract. Protection of software code against illegitimate modifications by its users is a pressing issue to many software developers. Many software-based mechanisms for protecting program code are too weak (e.g., they have single points of failure) or too expensive to apply (e.g., they incur heavy runtime performance penalty to the protected programs). In this paper, we present and explore a methodology that we believe can protect program integrity in a more tamper-resilient and flexible manner. Our approach is based on a distributed scheme, in which protection and tamper-resistance of program code is achieved, not by a single security module, but by a network of (smaller) security units that work together in the program. These security units, or guards, can be programmed to do certain tasks (checksumming the program code is one example) and a network of them can reinforce the protection of each other by creating mutual-protection. We have implemented a system for automating the process of installing guards into Win32 executables 1 . It is because our system operates on binaries that we are able to apply our protection mechanism to EXEs and DLLs. Experimental results show that memory space and runtime performance impacts incurred by guards can be kept very low (as explained later in the paper).

Towards Tamper Resistant Code Encryption: Practice and Experience

“…Our guards, which modify code depending on other code, offer several advantages over the software guards proposed by Chang and Attalah [4] and the those from Horne et al [9]:…”

“…If one would change a program at random places one could no longer guarantee the correct working of the application after modification. Several papers present the idea of self-verifying code [4,9] that is able to detect any changes to critical code. These schemes, however, do not protect against analysis of code.…”

“…Several years later, Chang et al [4] proposed a scheme based on software guards. Their protection scheme relies on a complex network of software guards which can mutually verify each other's integrity and that of the program's critical sections.…”

“…However, crashing is not very user-friendly. In the case of software guards [4,9], detection of tampering could be handled more technically by triggering another routine that for example exits the program after a random time, calls repair code that fixes the modified code (or a hybrid scheme, which involves both techniques), warns the owner about the malicious attempt through a hidden channel, . .…”

“…2. A partial function call graph containing recursive cycles: {f 4 , f 5 }, {f 4 , f 5 , f 6 }, and {f8, f9}. Functions f1, f2, and f3 are the functions giving control to the recursive cycles after decrypting all functions in the reachable cycles.…”