1. Secure Compiler Seminar 4/11 Visions toward a Secure Compiler Toshihiro YOSHINO tossy-2@yl.is.s.u-tokyo.ac.jp (D1, Yonezawa Lab.)

2. Talk Agenda Brief Introduction about TAL and PCC Introduction of my Master Thesis Visions toward a Secure Compiler

3. Brief Introduction about TAL and PCC

4. Background Program verification = Mathematically assure a program has certain properties Useful for security Memory access safety, information flow analysis, … Verifying low-level code directly reduces TCB TCB: Trusted Computing Base High-level code must be compiled after verified ⇒ We must trust the compiler Assemblers are much simpler than compilers

5. Current Techniques and Problems Code signing Based on public key cryptography Can prove the genuineness of code Cannot prove the safety by itself Signature matching Use a dictionary of malicious patterns and match target programs against it Employed in many antivirus systems Pass does NOT mean safety Often unable to detect very new virus

6. Proof-Carrying Code [Necula et al. 1997] Technique for safe execution of untrusted code Code consumer does not need to trust the producer Code distributed with the proof of its safety Producer creates a proof Consumer verifies the proof against his security policy

7. Proof-Carrying Code [Necula et al. 1997] Low consumer’s cost Consumer has only to verify the proof For example, by typechecking Tamper-proof If passed the check, code does NOT harm even if modified If modification makes the code fail the check, the code will not run and it is safe Otherwise code still obeys the consumer’s security policy

8. Typed Assembly Language [Morrisett et al. 1999] Extends a conventional assembly language with static type checking An instance of Proof-Carrying Code By type checking, it can guarantee Memory access safety Program never accesses outside the memory area allocated for it Interface consistency Type agreement of arguments / return value of functions etc.

9. TAL System Illustrated Type Checker Assembler Linker TAL System Code with type information Code Consumer

10. A Brief Example of TAL Program fact: movl%eax, %ecx movl$1, %eax loop: mull%ecx decl%ecx cmpl$0, %ecx jgloop end: {eax: B4} {eax: B4, ecx: B4} {eax: B4} Program Code (Same as conventional assembly languages) Type Information (Used to typechecking a program)

11. Related Work: TALK, TOS [Maeda, 2005] TALK: TAL for Kernel Morrisett et al. uses garbage collector for memory management in TAL For OS, GC cannot be assumed Must implement memory management (malloc/free) TOS: Typed Operating System An experimental OS written in TALK

12. Introduction of My Master Thesis

13. My Work for Master Thesis “A Framework Using a Common Language to Build Program Verifiers for Low-Level Languages” To help developers of program verifiers To be a common basis for verification of low-level programs Such as assembly and machine languages

14. Motivation: Verifiers are Hard to Develop Especially in low-level languages… Complex semantics Semantics of each instruction is complex There are many instructions in a language Low portability Low-level languages heavily depend on the underlying architecture Accordingly, entire verifier also depends on the underlying architecture

15. Our Idea Split a verifier into three parts 1. Design a common language, 2. Translate the target program into that language, and 3. Verify the translated program These parts are explicitly independent from each other Thus we can replace them easily

16. Our Idea Translated Program Target Program Translator Semantics of Common Language Result Success /Fail Verifier (1) (2) (3) Verification Logic

17. How Do We Solve the Problems? Coping with complex semantics Only translators care the semantics of the source language Translator is reusable Once description is done, we can reuse it Improving portability Verification logic is also reusable Once implemented, it can be used for other architectures simply by replacing translators

18. How Do We Solve the Problems? Translated Program Translator Semantics of Common Language Result Success /Fail Verifier Verification Logic Target Program Translator Program in Another Language

19. Overview of the Work Designed a framework to build program verifier Designed a common language ADL Discussed the correctness of translators Proved that the properties assured are preserved throughout translation Implemented the framework using Java

20. ADL: A Common Language Translated Program Target Program Translator Semantics of Common Language Result Success /Fail Verifier Verification Logic

21. ADL: A Common Language Design Concept ADL: Architecture Description Language From observation of many architectures Data is stored in registers and memory, and manipulates it according to program Only jumps are sufficient for control flow structure Expressiveness Arithmetics, logical operations, … C-like expressions Conservative semantics No need to describe indecent programs To simplify semantics

22. ADL: A Common Language Overview of the Language Imperative language which manipulates registers and memory 5 kinds of commands nop, error, assignment, goto, if-then-else Much like C than assembly Infix operators, parenthesized formulae Conditional execution by arbitrary condition using if command Only goto modifies control flow Unconditional branch

23. ADL: A Common Language A Brief Example data:... main: %ebx = &data; %eax = 0; goto &lp; lp: %eax = %eax + *[4](%ebx); %ebx = *[4](%ebx + 4); if %ebx == &null then goto &end else goto &lp; end: goto &end; data:... main: movl $data, %ebx movl $0, %eax lp: addl 0(%ebx), %eax movl 4(%ebx), %ebx cmpl $0, %ebx je end jmp lp end: jmp end ADLx86

24. ADL: A Common Language Restrictions ADL has a few restrictions by design Code and data are completely separated We assume NOTHING about memory layout of a program To simplify the semantics Some programs cannot be expressed However, most of decent programs can be written even under these restrictions To be discussed in the next slide

25. ADL: A Common Language > Restrictions Separation of Code and Data Do not treat code as data ADL programs cannot read / write code We cannot express the programs which uses dynamic code generation But, patterns of the generated code is fixed in many cases ⇒ Other solution is possible For example, prepare a function for each pattern of code

26. ADL: A Common Language > Restrictions Not Assume Memory Layout Casting is prohibited ADL distinguishes integers and pointers In real architectures, pointers are not distinguished from integers Pointer arithmetic is restricted Only pointer+integer, pointer-pointer are defined Other operations returns ‘undetermined’ Sufficient for array/structure operations and offset calculation

27. Program Translator Translated Program Target Program Translator Semantics of Common Language Result Success /Fail Verifier Verification Logic

28. Program Translator Translates low-level programs into ADL We must assure that program translators are correct Otherwise, we cannot trust the entire verifier Correctness is defined in the following discussion

29. State Program Translator What Is Correctness of Program Translation? Instruction = Function over machine states Correctness = Correspondence between states of two machines are preserved in translation Original Program Translated Program State State ’ State State ’

30. Program Translator How to Confirm Correctness of Translation Any programs result in corresponding states for any input ⇒ Correctness Total inspection is NOT realistic Theorem prover would be useful Automatic proving is one of future work But how to confirm the correctness of the description of the source language? At this time, we take empirical approach Test several cases using an interpreter

31. Verification Logic Translated Program Target Program Translator Semantics of Common Language Result Success /Fail Verifier Verification Logic

32. Verifies the properties of translated programs Function that takes a program and returns success or fail Soundness must be assured This is the task for the creator of a verification logic Here we do not discuss any further Definition: Soundness of a verification logic Verification logic V: State → Bool The set {S | V(S)} is closed about step execution If V(S), execution never falls into error state, and If V(S) and S→T (→ means step execution), then V(T)

33. Verification Logic Soundness of Verification Logic Machine States S such that V(S) Soundness = V(S) ∧ S→T then V(T)

34. Verification Logic Program Translation and Verification We proved the following theorem If program translator is correct, and Verification logic is sound, then ⇒ Verification on original program and translated program are equivalent Closed subset can be defined on the states of translation source language

35. Implementation Framework ADL data structures ADL interpreter Used to confirm the correctness of translators Translator, verification logic interfaces Translation rule compiler Compiles translation rule into Java implementation of a translator And for proof of concept, Translator from Intel x86 and SPARC A simple type checker

36. Related Works Foundational TAL [Crary, 2003] TAL type checker is still large TALx86 type checker consists of approx. 23k LoC in O’Caml (!) TCB is reduced by using a logical framework Designed a language called TALT on Twelf logical framework [Pfenning et al., 1999] Proved GC safety of TALT by machine Correspondence between TALT and realistic architectures are not discussed TALT type system is fixed Our work allows replacement of verification logics

37. Future Work Automatically confirm the correctness of translation Automatic testing Cooperating with emulators or debuggers Or, build a model and use a theorem prover Support dynamic memory allocation Currently all memory must be allocated statically Support concurrent programs Concurrency is not taken into consideration To apply for OSes, etc., concurrency takes an important role

38. Visions toward a Secure Compiler

39. What Is Secure Compiler? A compiler which produces certified code For example, TAL code as output Like Popcorn compiler in TALx86 Safe dialect of C → TALx86 A compiler which assures correct compilation (optionally) Like credible compiler [Rinard, 1999] Reduces TCB

40. Motivation Infrastructure has been built TALK, TOS [Maeda, 2005] Verifier framework [Yoshino, 2006] Next we have to build a house on it! Most people do not want to write low- level code directly ⇒ Secure Compiler

41. Toward Secure World If we built a secure compiler… Memory-error-free systems Prevent memory-error-based attacks OS kernel, core libraries, network server… Writing secure code Vulnerable code will result in verification failure So code security will be improved Rest to be discovered…

42. Tasks to Do Determine what properties to assure Memory access safety? Information flow? Must be mechanically checkable Design the verification logic Use verifier framework? Design the language Target: TAL-base? ADL? ADL can be used as certified language Register allocation is done, so simple mapping will be possible… Source: ???