CS261 Projects

General information

Your term project should address a research issue in computer security and consist of the design of some computer security system or technique, or the analysis and possible improvement of some existing system or technique. The main goal of the project is to do original research on a problem of interest in computer security.

You should work in a small group; I expect that teams of approximately 2--3 will be appropriate for most projects. Of course, expectations will be adjusted according to the number of people in your group. I will not categorically rule out solo teams, but I expect that working in groups will allow you to tackle more substantial research issues. If you have trouble finding a project partner, I can help you get matched up with someone else by maintaining a list of people seeking teammates.

Projects will be evaluated on the quality of their research in computer security. At the end of the semester, you will write a conference-style paper on your work. See below for more details.

I expect that most projects will fall (more or less) into one of two categories:

1. Design. Design projects will usually attempt to solve some interesting problem by proposing a design; implementing a prototype; and using the implementation as a basis for evaluating the proposed system architecture.
2. Analysis. Analysis projects might, for example, study some previously-proposed implementation technique, existing system, or class of systems; evaluate its security properties; find flaws, or strengths, in it; and provide new insight into how to build secure systems.

If you're at a loss for a project topic, I've prepared a list of possible project topics that you can peruse as examples of how to a pick a suitable project. See below. But don't feel limited to these suggestions! They are intended only as examples.

You're welcome to come discuss possible project ideas with me, if you like. I'm happy to make myself available to discuss projects.

A final suggestion: Aim high! The top projects could lead to publication. In past years, the best two or three projects have consistently led to publications.

The process

You will write a concise (approximately 1 page) project proposal that should clearly state the problem you will be solving, the key challenges for new research, your plan of attack (including milestones and dates). If there are any special resources you might need from me, mention this as well. You might mention any relevant papers that you are aware of.

The project proposal is due Monday October 15.

Here's how to submit your proposal. You should put together a web page for your project; currently all it needs to contain is the project members, their email addresses, title of your project, and proposal. Then just email the URL for your project web page to daw@cs.berkeley.edu by Oct 15th.

In mid-November I might ask you to write a concise status report so I can make sure the projects are on-track. I am always available to meet with any groups who would like to discuss their project, request additional resources, or ask for advice.

You will also be required to present your project at a poster session, to be held on Dec 5th, 4-6pm, in Woz lounge.

Finally, a project report will be due on Monday, Dec 17th, at 9am. No exceptions or extensions will be granted, so get it in on time! See below for instructions.

The final report

You are expected to write a technical paper, in the style of a conference submission, on the research you have done. State the problem you're solving, motivate why it is an important or interesting problem, present your research thoroughly and clearly, compare to any related work that may exist, summarize your research contributions, and draw whatever conclusions may be appropriate. There is no page limit (either minimum or maximum), and reports will be evaluated on technical content (not on length), but I expect most project reports will probably be between 7--15 pages long.

If you are not familiar with writing conference-style papers in computer science (or even if you are), the following resources may help:

• Advice on research and writing
• More resources on writing systems papers
• Detailed advice on technical writing
• Armando Fox's hints on writing good systems papers

You may submit your project report electronically or on paper. I prefer electronic submission, although you may choose either. In either case, the deadline is the same.

If you submit electronically:

1. It must be in a format which is easily readable on Unix platforms: that means HTML, Postscript, or PDF is fine (but not Microsoft Word).
2. Place a link to the file on your project web page (see here for the list of project web pages), and send me email with the URL. I will send you confirmation of receipt.

Example ideas for project topics

Note: Many of these examples are quite generic; be sure to narrow down the topic substantially and propose something concrete and focused. If you are interested in any of the project topics below, feel free to talk to me about it; I may be able to make some more concrete suggestions.

Analysis and attacks

Security analysis of privilege-separated applications

Pick a privilege-separated application (e.g., OpenSSH, vsftpd, qmail, Postfix, Openwall popa3d, Openwall telnetd, OpenBSD ntpd, OpenBSD syslogd). Do a security analysis of the implementation and the effectiveness of the privilege separation. If someone compromises the unprivileged portion, how much harm can they do? Any suggestions for what would improve the security of the application, or make it easier to perform security reviews of the code?

Security auditing

Audit a widely-used and under-scrutinized open-source package that is security-critical. Report on your experiences and lessons. How would you re-structure/re-implement the system to make it more robust? What tools would have made your auditing task easier? How effective are existing tools?

Security review of published schemes

Pick any recently published paper that proposes a new security mechanism or scheme. Ask the authors for the code. Perform a careful security review of the paper's scheme; does it meet the claims made for it? To find recent papers, you could peruse recent proceedings of Usenix Security, IEEE Security & Privacy, ACM CCS, ISOC Network and Distributed System Security, or other security-related conferences.

Fuzz testing

Fuzz testing has recently drawn interest because it is very cheap, and more effective than you might expect (given its low cost) at finding some kinds of security bugs. There a number of fuzzing-related projects you might consider.

(1) Apply fuzz testing tools to several different OSs. Can you draw any conclusions about how they compare with regards to their robustness? For instance, OpenBSD is a BSD-based distribution that is known for its focus on security; you might compare how OpenBSD compares to FreeBSD/NetBSD as measured by fuzz testing. (See also this research from U Wisconsin.)

(2) Devise and perform an experiment to characterize the tradeoffs between fuzz testing and static source code analysis. You might select several applications, apply both fuzz testing and a source code analysis tool to those applications, and compare the two techniques in terms of how many vulnerabilities they found, how many false positives they generated, and the time or effort it took to apply these techniques. (For static analysis tools, you might consider Fortify SCA, Coverity, or FindBugs, depending upon what kinds of applications you decide to analyze.)

(3) White-box fuzz testing: In standard fuzz testing, one runs the programs on random inputs and declares a testcase to have failed if it causes the program to crash or freeze. Thus classical fuzz-testing is a purely blackbox testing technique. Can you devise a more discriminating oracle for classifying testcases as passing or failing, perhaps by examining the program's state as it executes? For instance, does compiling the program with a bounds-checking compiler (like CRED), running it under a runtime analysis tool (like Valgrind memcheck), and/or with a debugging library (e.g., debugging malloc, ElectricFence) help fuzz testing to find more security holes?

(4) Conventionally, fuzz testing is normally done using end-to-end blackbox tests. However, one could imagine performing unit testing using fuzzing methods. Is there any way to make this effective? The obvious problem is that such tests might generate many false alarms, if the randomly generated tests violate the module's preconditions or otherwise cannot arise during normal program execution. Can this problem be surmounted?

(5) Fuzz testing of web browsers: There have been some recent advances in developing fuzzers to test web browsers. However, so far the tools have focused primarily on fuzzing a few core formats (e.g., HTML, Javascript), but web browsers support many many kinds of web objects. You might consider picking a format (e.g., CSS, Flash, ActionScript, SVG, PDF, XSLT, Quicktime, Shockwave, various audio/video formats) that is widely supported but hasn't been explored as well. References: mangleme (only fuzzes HTML; doesn't test CSS style sheets, Javascript, etc.); jsfunfuzz (fuzzes Javascript; has found hundreds of bugs in Firefox).

Web security

CSRF vulnerability scanning

Cross-request forgery (CSRF) vulnerabilities have been called a sleeping giant of web security -- partially because current web vulnerability scanners do not detect these vulnerabilities, and because it is suspected that these kinds of vulnerabilities are rampant. Develop a way to automatically scan web sites for cross-request forgery (CSRF) vulnerabilities. Implement it (perhaps as a plugin to a popular vulnerability scanner like Nessus or Nikto or somesuch). Try it out: how well does it work?

Ajax security

Develop a way to detect vulnerabilities associated with Ajax and client-side Javascript. For instance, many web services split the code between the server and the client: some of the code runs on the server side (e.g., as Java or PHP), and some of it runs on the web browser (e.g., as Javascript provided as part of the web page). Develop methods to detect vulnerabilities in these kinds of web services. For instance, one common vulnerability involves executing input validation code or other trusted logic on the client instead of on the server; the problem is that the client may be under the attacker's control and so the attacker may be able to bypass that code.

Or, propose a tool or framework for building secure Ajax services.

HTML filtering

Work out the details of how to build a robust, functional HTML filter (along the lines of the homework, but exploiting what you have learned). How much of the web can one still support without endangering security overly much?

Software security

Efficient taint-tracking for Java

Most prior work on runtime taint-tracking has focused on languages like C (or, in one case, PHP). Study techniques for runtime taint-tracking of Java. You may want to focus on only tracking taint for Strings, as that is enough for some applications of taint-tracking. Minimizing the performance overhead is important for many applications. How efficient can you make your solution? Can you handle issues such as reflection, dynamic class-loading, and multi-threading? You might want to consider building on existing tools for bytecode instrumentation/transformation or aspect-oriented programming. Also you might consider performing some kind of static analysis (either of the JVML bytecode or the Java source code) to determine which objects need to be instrumented and which ones can never be tainted and thus do not need to be instrumented, if this reduces the runtime overhead. What is the performance cost of byte-level taint tracking (where you track the taint status of each byte of the String independently) vs object-level taint tracking?

One intriguing application of taint-tracking for Java includes automatic detection of SQL injection vulnerabilities, cross-site scripting vulnerabilities, and other security holes in web services (see the Livshits and Lam paper you read for some examples). You could try implementing this and seeing how effective it is.

Inferring security annotations for C/C++ programs

Microsoft has proposed SAL, a set of annotations for C and C++ code intended to help avoid buffer overrun and similar vulnerabilities: the programmer annotates their code with information about buffer lengths and the like, and a static analysis tool checks those annotations to detect possible bugs. See also Deputy, an open-source system from Berkeley with similar goals. Writing these annotations can get a bit tedious. Can you design a dynamic analysis tool that observes code as it runs and infers SAL/Deputy annotations from how the code is used? For instance, if you run the program on 1000 inputs, and in every case, function f() is only called with null-terminated strings, you might infer a SAL/Deputy annotating asserting that the argument to f() is always null-terminated. (See also Daikon, though Daikon is a general-purpose tool; you can probably do a lot better by focusing specifically on the kinds of properties that SAL/Deputy are designed for.) Tools like Valgrind, CIL, ltrace, etc. could be a good building block for this project.

Avoiding integer overflow vulnerabilities in Java

Modern type-safe languages like Java improve on C/C++ by eliminating many classes of vulnerabilities that have plagued C/C++ programs, but it does not eliminate integer overflow and integer casting vulnerabilities. One way to write Java code that is free of integer overflow bugs is to use java.lang.BigInteger (infinite-precision integers) instead of int. However, this introduces a notational burden and makes code less readable: e.g., instead of writing x*(y+z) - 2, the programmer must write x.times(y.plus(z)).minus(new BigInteger(2)). Can you come up with an extension to Java that eliminates this notational burden? For instance, you might build a tool that performs source-to-source translation to transform the former short version into the latter long version before compilation, allowing programmers to use the ordinary binary operators with BigIntegers.

Detecting algorithmic denial-of-service vulnerabilities

An algorithmic denial-of-service vulnerability is a flaw that allows an attacker to provide a carefully chosen input that will trigger worst-case performance behavior in the application under attack and deny service to legitimate users. For instance, at one point the Apache web server had a flaw where URL parsing took time quadratic in the number of "/" characters in the URL; by sending a URL consisting of thousands of slashes, one could cause Apache to consume a huge amount of CPU time. Develop a method to automatically detect these kinds of bugs in application source code. For instance, you might consider using Trend-prof (see also this paper) to detect performance anomalies, and try to infer attacks.

Privilege separation

Pick an interesting security-critical application that is currently implemented monolithically. Investigate how to apply privilege separation techniques to reduce the size of the TCB. Can you implement your new design and evaluate its effectiveness? If you were designing the application from scratch, how would you design it to maximize its security? Any network client or server could be a good candidate for investigation.

Improving Plash

Improve the security of Plash by eliminating the possibility for the sandboxed process to issue syscalls directly and bypass the Plash monitor. For instance, right now, Plash imposes no limits on the ability of the sandboxed program to use the network. Propose and implement some solution to fix this. (seccomp? ptrace? SFI? something else?) Here's a list of Limitations of Plash, if you want to have some ideas on how it could be improved.

Software verification tools for security

Researchers have recently made dramatic improvements in tools for software verification. See, e.g., ESC/Java2 and JML (verification tools for Java) and Spec# (a verification tool for C#). These tools allow programmers to verify properties, such as that the program will never throw a NullPointerException, ArrayIndexBoundsOutofBoundsException, or other runtime exception, and that the program will never use uninitialized memory. Since unexpected exceptions can cause surprising behavior (which is dangerous in a security-critical program), these could be useful for secure programming. You might study how useful and expensive these are for security. For instance, pick one or two security-critical applications written in Java or C# and attempt to verify that they satisfy some useful property (e.g., free of runtime exceptions). How many annotations did you have to add? How much time or effort did it take? Did the effort reveal any vulnerabilities? Or, you might try to see if you can express any of the security requirements for those applications into the JML/Spec# modelling language and see whether it is possible to verify that those requirements are met by the code. Are the JML/Spec# annotation languages rich enough to specify important security policies? If not, what kinds of extensions would be useful for security? Are the existing tools powerful enough to verify that code meets those requirements?

Tools for vulnerability detection

Study ways to build tools to help automate the process of reviewing security-critical applications. Can you use runtime testing, static analysis, model checking, formal verification, or other techniques to detect any interesting classes of common security holes? You may want to consult CWE's list of top classes of vulnerabilities for the past five years.

Virtual machines for security

Virtual machines (e.g., VMWare, Xen, etc.) seem to provide a powerful mechanism for executing dangerous actions in an isolated environment. How secure are existing commercial virtual machines? How hard would it be for a malicious guest application to defeat the VM and harm the host OS? (For instance, an interesting project would be to look at a few commercial or open source VMs and try to see if you can break isolation.) How would you design a high-assurance VM to be sure that isolation could not be compromised and to make it as easy as possible to verify that these security goals have been met?

Miscellaneous

Defenses against phishing

There has been a great deal of research on defenses against phishing. Can you improve upon any of them?

Verifiable distributed computation

The Internet is a vast resource of idle machines; we might like to harness these spare CPU cycles by offloading our lengthy computations to other computers. But in any such distributed setting, how do we know that the result that comes back is the correct one we wanted? Can we do anything?

Formal modelling of security systems

Build a formal model of some aspect of a security system, and rigorously evaluate its properties. For instance, you might look at the state machine associated with a TCP/IP stack, and model how the various network events can affect the state. You could build a formal model of actual behavior by working from the OS source code or by exhaustively testing the possibilities. Then you might build a formal model of intended behavior -- e.g., by working from the RFC -- and you would check whether the specification matches the verification. Or, you might build a second model from a second operating system, compare where their behavior differs, and study whether this has any consequences for how to write portable security code. Such models might also be useful for intrusion detection as well.