Managing Realeyes Memory

I have been at work on a new project that I hope to announce soon. But at the moment I need a break to let the algorithm for a particularly tricky function to germinate. So I am going to describe how I do memory management in the Realeyes IDS.

First, I have to say that this is not a generic memory manager. It is specific to my application. It may be possible to adapt it to other applications, but the key word here is 'adapt'. However, it will hopefully give anyone who is considering doing their own memory management some food for thought.

The reason I do my own memory management is to avoid fragmentation. The Realeyes IDS manages a lot of sessions simultaneously, so memory has to be used efficiently as possible. If a buffer were allocated for exactly the size of a packet's data, the overall buffer space would develop lots of pockets of unusable space. But to set the size of data buffers to the largest allowed by the Internet Protocol would also be inefficient, because there are a huge number of tinygrams in network traffic.

The solution is a compromise. I allocate fixed size buffers in sizes that are designed to waste as little space as possible. The smallest is 64 bytes. The next is 105 bytes, and the next, 128 bytes. So if 56 bytes are needed, the first size buffer is allocated. If 65 bytes are needed, the second. And if 108 bytes are needed, the third.

If you did a double take on the 105 byte buffer size, there is a method to my madness. These buffers are kept in pools of 8 Kilobytes. A 64 byte buffer pool will hold 256 buffers, and a 128 byte buffer pool will hold 64 buffers. Both of these will fit exactly into the 8K pool with no wasted space. To fine tune this a bit, I found the buffer size between them that wastes the least space. Allocating 78 buffers of 105 bytes each uses 8190 bytes, which wastes 2 bytes of space in an 8K pool.

Here is the complete list of buffer sizes: 64, 105, 128, 195, 256, 390, 512, 744, 1024, 2048, 4096, 8192. If larger sizes are needed, multiple adjacent pools are allocated, up to 64K. Again, this is specific to the Realeyes IDS application, which can guarantee that no buffer larger than 64K will ever be requested.

When the application is initialized, a huge buffer (many megabytes) is allocated and it is divided into 8K pools. Then, when a buffer is requested, if there is no pool for the appropriate buffer size already selected, the next available pool is assigned to provide buffers of that size only, and the first buffer in the pool is returned to the requester. If a pool already exists for the buffer size and has free buffers, a buffer from that pool is returned.

To handle the requests, each allocated pool is kept on one of three queues for that buffer size. The reason there are three is that the entire queue must be locked while buffers are being allocated or freed. The Realeyes IDS functions to handle semaphores allow for a caller to request an immediate return if a lock is already held. This means that if one of the pool queues is in use, the caller can try the next one. The rae_mem_mgmt.c module keeps track of the last queue accessed and uses a simple round robin method to check the next queue.

So far, so good. But there are still some loose ends. What if all of the buffers in a pool are in use, and that pool is at the head of the pool queue? For that matter, what if the first 1,000 pools in a queue have no available buffers? This is where the manager comes in.

For each pool size there are Full and Free queues. The manager periodically (about 500 times a second) checks each of the available queues, removes all pools that have no available buffers, and puts them on the Full queue for that buffer size. Pools on the Full queue that have available buffers are put on the Free queue. And pools that are on the Free queue are divvied up so that each available queue has approximately the same number of pools.

There are a few other steps in managing the queues, which is done in rae_mem_ctl.c. If an allocated pool has all of its buffers freed, it is put on a general available queue to be reused for possibly a different buffer size. Also, there is a queue at each buffer size for full pools that have not had available buffers for a period of time. This is only checked once a second to see if buffers have been freed.

So does it work? In the pilot project I have been running for over a year, the statistics I have collected show that the average number of sessions being managed simultaneously is around 20,000. Assuming an average size of 16K per session, that is 325M of data, plus the application data about each session. And that is just a snapshot. There are many Gigabytes of data being examined over the course of an hour. When the IDS does run out of buffers (I'm working on it, OK?!?), it is because the application hasn't released them when it should, not because the memory management is bogging down.

So that's the essence of how memory management is handled in the Realeyes IDS. However, because the application uses multiple processes instead of threads, the memory must be shared. I will cover that in a future post.

Later . . . Jim

Handling Semaphores in C Programs

A while back, Carlo Schroder, over at LinuxToday.com, put out a request for articles on programming. Now that I have put the downloads for Realeyes IDS version 0.9.5 up on SourceForge, I get to have some fun answering her call.

What I have found in programming books, including those by the late W. Richard Stevens (which I turn to most often) is usually a good start, but never the whole story. But since this is not a general programming text, I will focus on a single issue in detail. This post will cover using semaphores.

First, a little background on locks, in case you have never used them. In *nix, process locks are implemented with the semaphore system calls. Since I use child processes that share memory, I have to implement semaphores. Threads use pthread_mutex calls, which do essentially what these functions do, and then some.

The most common reason for implementing locks is if you have multiple concurrently running processes or threads that have access to the same variable in memory. Obtaining or releasing a lock is guaranteed by the operating system to be completed without interruption. A single line of C code, such as

 if (flag & 4)

requires a minimum of two machine instructions:

• Get the value of flag from memory into a register


• Compare the value to zero
  

It is possible the thread running that code could be swapped out after getting the value, but before comparing it, and the value of flag could be changed by another thread, making the comparison invalid. By requiring every thread to get the flag lock before reading or writing the value of flag, only one thread accesses flag at a time.

The rule of thumb for the code while a lock is held is to do only what requires holding the lock, and no more. Often there are less than ten instructions between getting and releasing a lock. However, sometimes there are a couple dozen, because all of them require holding the lock. My memory management is an example of this, and I will try to cover it down the road.

OK, now for how I implement semaphores. For some reason, the caller is required to define the semun structure. This definition is taken from the semctl man page and is in the rae_lock_mgmt.h file.

union semun {
 int                val;    /* Value for SETVAL */
 struct semid_ds    *buf;   /* Buffer for IPC_STAT, IPC_SET */
 unsigned short int *array; /* Array for GETALL, SETALL */
 struct seminfo     *__buf; /* Buffer for IPC_INFO */
};

All of the following code is from the rae_lock_mgmt.c file. First, I define a global to keep track of a held lock. This is done so that if there is an interrupt, such as a segmentation fault, while the lock is held, it can be released in the signal handler by calling rae_release_lock. The caller must set these fields to the address of the variables used to track the semaphore ID and index.

/* Pointer to currently held lock */
int *rae_held_lock = NULL;
int *rae_hl_index = NULL;

Before a lock can be used, it must be initialized. I use a lot of locks and have found that different Linux distributions have different defaults for the maximum number of semaphores an application may allocate. To keep the number of allocations down, I have grouped the locks by functionality, and each group gets a semaphore set (or array) which only uses a single semaphore ID. Therefore, the number of locks in the group is passed to the init function.

int rae_init_lock(int il_size)
{
 int i, il_semid = -1;
 union semun sem_union;

The semget call returns a new semaphore set. Then each lock in the array is initialized to a value of 1 to indicate that it is available. If that fails, the semaphore ID is set to a negative value, which is -1 if the semaphore set is released, and -2 if it is not.

  if ((il_semid = semget(IPC_PRIVATE, il_size, (0660|IPC_CREAT))) > 0)
 {
   sem_union.val = 1;
   for (i=0; i < il_size; i++)
   {
     if (semctl(il_semid, i, SETVAL, sem_union) == -1)
     {
       if (semctl(il_semid, 0, IPC_RMID, sem_union) == -1)
         il_semid = -2;
       else
         il_semid = -1;
     }
   }
 }
 return(il_semid);
} /* end rae_init_lock */

When the application shuts down, the locks must be freed. If not, they remain allocated by the system. You can run 'ipcs -s' to see the locks that are held. If an application fails to relaese a lock, you can run ipcrm (read the man page) as root to release it.

Notice that the location of the semaphore ID in the application is passed to this function, and it is set to zero by the function. This is because the sem_ctl command, IPC_RMID, ignores the index and simply removes the entire semaphore. Also, I prefer to do as much as possible in a function so the caller does not have to worry about the details. This way, when I call the same function from different places, I reduce the risk of forgetting to set something.

int rae_free_lock(int *fl_semid, int fl_idx)
{
 int fl_stat = 0;
 union semun sem_union;

 if (*fl_semid == 0)
   goto out;
 sem_union.val = 1;
   fl_stat = semctl(*fl_semid, fl_idx, IPC_RMID, sem_union);
 *fl_semid = 0;
out:
 return(fl_stat);
} /* end rae_free_lock */

When a lock is needed for a memory location, the get lock function is called with the lock identifier, which consists of the semaphore ID and the index in its array. I have added a wait flag to allow some locks to be conditional. In my memory management code, I have three available buffer queues, and if the lock on one is held, the caller can simply try the next one without waiting.

The semop call gets and releases a lock by subtracting the supplied value from the semaphore's value. What this means is that you, the programmer, are responsible for defining what value indicates a held or released lock. By keeping all of this logic in a pair of functions, you have control over how it is implemented. All of the examples I have seen use 1 to indicate the lock is available, and 0 to indicate that it is held. I can imagine how other values might be used, but it seems ridiculously complicated and prone to error.

int rae_get_lock(int gl_wait, int gl_semid, int gl_idx)
{
 int gl_stat = 0;
 struct sembuf sem_b;

 if (gl_semid == 0)
 {
   gl_stat = -2;
   goto out;
 }

The semaphore buffer structure is defined in the system headers. This is where the semaphore array index and operation are set. And as I said before, this function allows the caller to wait or not, which is accomplished by using the semaphore flag bit, IPC_NOWAIT.

The SEM_UNDO flag bit is supposed to reverse the operation when the process terminates, which implies that if the process fails while holding a lock, the system will release it (but not free it). However, in my experience, that doesn't always work, so I have included the capability to do this in my interrupt handlers, as I mentioned above.

  sem_b.sem_num = (short int) gl_idx;
 sem_b.sem_op = -1;
 if (gl_wait & raeLOCK_WAIT)
   sem_b.sem_flg = SEM_UNDO;
 else if (gl_wait & raeLOCK_NWAIT)
   sem_b.sem_flg = SEM_UNDO | IPC_NOWAIT;
 else
 {
   gl_stat = -1;
   goto out;
 }

This is the heart of the function. Read the semop man page for a lot more detail, but the general idea is as follows. The semop system call will attempt to subtract 1 from the current value of the lock. If that value is 1, the operation occurs immediately. Otherwise, the call will wait or return with the errno value of EAGAIN if the lock is held. Of course, there is the possibility the call will fail entirely, which must be handled.

If the lock value is set to 0, this means the lock is obtained, and this function sets the semaphore ID and index in the global lock tracking variables.

  if ((gl_stat = semop(gl_semid, &sem_b, 1)) == -1)
 {
   if ((gl_wait & raeLOCK_NWAIT) && errno == EAGAIN)
     gl_stat = 1;
   else if (errno == EIDRM)
     gl_stat = -2;
 }
 if (!gl_stat && rae_held_lock != NULL)
 {
   *rae_held_lock = gl_semid;
   *rae_hl_index = gl_idx;
 }
out:
 return(gl_stat);
} /* end rae_get_lock */

This is the reverse of the get lock function, in that it adds one to the lock value. There is no wait flag for releasing a lock, so only the semaphore ID and index are supplied. If the lock value is set to 1, the lock is released and this function clears the semaphore ID and index in the global lock tracking variables.

int rae_release_lock(int rl_semid, int rl_idx)
{
 int rl_stat = 0;
 struct sembuf sem_b;

 sem_b.sem_num = (short int) rl_idx;
 sem_b.sem_op = 1;
 sem_b.sem_flg = SEM_UNDO;
 rl_stat = semop(rl_semid, &sem_b, 1);
 if (!rl_stat && rae_held_lock != NULL)
 {
   *rae_held_lock = 0;
   *rae_hl_index = 0;
 }
 return(rl_stat);
} /* end rae_release_lock */

This set of functions makes using semaphores as easy as:

• Init lock


• Get lock


• Release lock


• Free lock
  

Of course, the caller code must be well thought out to prevent a deadly embrace. That is accomplished by keeping the code using the Get and Release calls as simple as possible, and making sure the instructions between them absolutely require the lock.

Later . . . Jim

Realeyes IDS 0.9.5 Released

There is a new download available for the Realeyes IDS version 0.9.5. Read the release notes for details, but essentially, this release is about new user interface features. The screenshots have been updated and there are two new demos, one on installation and configuration, and the other on the new features. There are links to all of this on the technology page

The college where I have been running a pilot project was able to provide an upgraded host system. This led to the discovery of a couple of issues with the IDS sensor software that I had not seen before. One was the signal handler for user interrupts, and the other was in releasing locks if shutting down. Both of these are fixed, although I might make some further improvements before the next release.

The big news is that the college security software faculty heard about the pilot project and is going to use Realeyes in the curriculum. This means that I am now running the system in a lab environment. And all of this means it gets tested more thoroughly. I did find a couple configurations that would not run successfully, so hopefully, the installation process is more solid.

I am hoping that this will help build the community around the project. I have had some interest expressed in working on code, as well as testing it in other environments. So it is moving in the right direction, and with a little luck and some elbow grease, the next release will be 1.0.

Later . . . Jim

Command Line vs. GUI Reality Check

I was reading Kyle Rankin and Bill Childer's Point/Counterpoint column on "Mutt vs. Thunderbird" in Linux Journal over breakfast the other day. It mostly boiled down to the perennial text vs. graphical user interface argument. And since I don't have a strong opinion about either mail client, it got me to thinking about the real difference between the two interfaces.

Before I dive into the firepit, let me explain that I spent the first half of my career in the IBM mainframe world, which meant writing Job Control Language (JCL) to submit batch jobs. Part of that time, I spent maintaining a TCP/IP stack written in assembler language. Compared to that, the distance between the command line and the GUI is much smaller than their advocates seem to realize.

To be honest, I didn't use GUIs for a quite a while after they became available. I didn't fight them so much as I just didn't find them intuitive or efficient. However, as usability studies improved the interfaces and processor speeds increased their capabilities, I gradually came to appreciate what they provide.

I think that much of the bias in favor of GUIs comes from the saying, "A picture is worth a thousand words." And certainly, pictures of cityscapes or family reunions or battlefields convey much more information than even the best description. However, imagine having to draw a picture to say, "I'm thirsty," or, "Turn left at the next light," or even to describe how to use a graphical user interface to a first time user. Very often, a few words are more efficient than any picture.

Obviously, each interface has its strengths and weaknesses. So what I was wondering, as I read Bill and Kyle, was whether it is possible to generalize the strengths of each. I'm not suggesting that I have it completely figured out, but I think there are, in fact, some common elements for the situations in which each is superior to the other.

To begin with, I am assuming that the interface in question has gone through enough testing and real-world use to be very good at what it is supposed to do. The Linux/Unix command line's many utilities that can be piped or scripted to create miniature applications is the best example of that interface. And I prefer the GNU collection because of the extra arguments that have been added by developers scratching an itch. Meanwhile, the current state of the art web browsers are good examples of GUIs.

Command line interfaces are best for data that is concise and reasonably well known in advance. For example, comparing the dates of two copies of the same file in different directories:

ls -l /a/b/c/file /d/e/f/file

Or finding information about running processes:

ps aux | grep program

Or starting and stopping daemons:

/etc/init.d/daemon start
/etc/init.d/daemon stop

The last two examples are mostly used by administrators, and I think that the command line is especially appropriate for that type of work. However, as a developer, I find that the most repetitive tasks I perform are done so most efficiently on the command line. And I personally almost never use a file browser, preferring the flexibility of the command line.

Graphical user interfaces are best for displaying somewhat large amounts of data that needs to be analyzed to determine what to do next. When a web page is loaded in a browser, it is usually not known in advance what link(s) will be selected. A window with a helpful design to guide the user to the proper selection and a mouse pointer that can be moved directly to the appropriate target is a highly effective use of this interface.

With this in mind, I started thinking about the use of both interfaces in the Realeyes IDS.

The installation is performed by executing several shell scripts on the command line. Many of the responses have a default, which allows the Enter key to be pressed for a majority of the responses. And where that is not possible, simple information, such as an IP address or a single name, is entered. To begin with, a graphical environment is not required. And second, there is no switching between the keyboard and the mouse. This makes deploying a new sensor or updating the application very fast after the first couple of times.

In contrast, the analysis window displays large amounts of information that may not be selected in a sequential order. Having the data in a scrollable list allows me to focus on one column, while watching another with my peripheral vision. This allows me to see patterns that might not be apparent in a text display.

Another advantage of the GUI comes when I want to view the session data. This is done by right clicking on a report and selecting Playback from the popup menu displayed. When the reports are sorted by Event, all of those with the same Event are similar enough that a quick glance at the session data is sufficient to determine whether it warrants more attention. Then a single click closes the window. This means that I can often rip through reports, spending only 2 or 3 seconds on most of them.

The GUI also provides tabs that allow me to search for trends or get summary reports without losing my focus on the original display. And the status frames automatically notify me if there is something that needs attention, without my having to query for it.

There are some administrative functions that could be performed from the command line as easily as the GUI, and in the early versions of the Realeyes IDS, it was the only way to do that. However, having them incorporated into the GUI is very convenient.

The downside of this is a lack of flexibility. In order for a capability to be available, there must be code in the GUI application. The command line gives an administrator complete control of maintenance procedures, and under certain circumstances, this is the only option.

From a design perspective, the choice of command line vs. GUI seems pretty straightforward. First, how quickly does the code need to be produced? Second, which interface makes the user most productive? While there is plenty of room for different points of view on the answers to these questions, it is simply not true that one is always better than the other.

Later . . . Jim

Database Security

I saw the following on the webappsec list at Security Focus:

| I've heard this preached before.
|
| Using JDBC properly can help protect against SQL Injection.
|
| What protections does JDBC provide?
|
| Does java encode the input to not be malicious?
|
| I'm curious where in the java source/libraries does jdbc help
| to mitigate malicious input when using jdbc.
|
This preach is applicable for any programming language. It
all depends on how well you have done input & output
validation. As in what input you expect & what input is
malicious for your app. If all goes well you can make SQL
injection very difficult or even impossible . The reason I
say difficult, because it all depends on how well the SQL
injection is crafted. As far as I recollect I don't think
JDBC or for that case even java gives you predefined class
for doing that. But there is quite a possibility that some
one on the internet must have surely written these classes.
--
Taufiq
http://www.niiconsulting.com/products/iso_toolkit.html

I don't disagree with Taufiq's assessment. However, I do disagree with his acceptance of the status quo. I wrote a rant on this blog responding to a complaint that security professionals are not taken seriously. In it, I pointed out that the security industry should promote improving the security climate, not just react to it with solutions 'for a price'. The example I gave was *DBC libraries.

The JDBC package, java.sql, does not supply any security parsing. This is not the real workhorse, but it should at least provide a method for this. Each database supplies a jar that java.sql classes call to access the specific database. This is where security parsing must be handled.

The thing is that parsing input is tricky. The first step is to validate that the input is correct for the column data type. This is reasonably straightforward for simple types like integer and varchar. But the way different databases support binary data and very large fields is not consistent. There is also support for non-standard data, such as PostgreSQL's support for the inet data type.

The JDBC Connection interface includes the getMetaData method, which returns the information supplied by the specific database library, some of which is unique to that database. There are not only differences between databases, it is even possible that there are differences between versions of the same database. This could be an issue for an application because:

Some DatabaseMetaData methods return lists of information in the form of ResultSet objects. Regular ResultSet methods, such as getString and getInt, can be used to retrieve the data from these ResultSet objects.

All unique information must be verified for every version of the database supported. And if you are supporting multiple databases, it is that much more difficult.

The next step is to escape all characters that have special meaning, such as single quote and backslash. But again, each database has its own special characters that must be accounted for, such as ampersand in Oracle, and the E'...' escape pattern in PostgreSQL.

Update: Eric Kerin points out in his comments that the PreparedStatement interface does this, and after some testing I have found that this is the case. My excuse is that there is nothing in the javadoc for the SQL package or the PreparedStatement interface that explains this. Instead the documentation promotes it for optimizing frequently used statements. See my reply below for further responses to comments.

Also, there is a good article on this issue at the Open Web Application Security Project, which I found by googling for java and "sql injection".


The current situation places the responsibility for security on the thousands of application programmers, who must now dig into the internals of the database(s!) on the backend of their applications. If instead, the database development teams provided a parser for each field of data, it would be possible to determine if the input would result in a message, like this one that I was able to create from testing various input sequences:

WARNING: nonstandard use of \' in a string literal

ERROR: relation "foobar" does not exist

STATEMENT: select foo from foobar

I'm still working on parsing that construct and reworking it in a way that does not reject the data out of hand, because it might be a legitimate description of an Event, or possibly a Trigger definition. I am fortunate, because the input is not read from the network. You might not be so lucky.

And before I leave this topic/rant, I must point out that application programmers need to work closely with their DBAs to be sure that permissions are set on tables to allow only as much access as absolutely necessary and no more. If you don't have a DBA and/or maintain the database yourself, you need to become very familiar with the levels of GRANTing access and the use of roles to at least limit the damage when SQL injection attacks succeed. In my own experience, as well as reports from others, the attacks on applications, and databases especially, is continuing to increase.

If anyone is interested in the database security in the Realeyes UI, check out the Database and ValidatorDBForm modules, and then see how they are used in any of the Window*Base.java modules at the UI subversion repository. The ValidatorDBForm class includes the InterfaceAdminExtendedProcessing interface to do extra contextual error checking, which really is the job of the application. There are some pretty good examples of its use in the WindowsRules(Triggers/Actions/Events).java modules.

I'm pretty sure I'm talking for all application developers when I say (as a security guy), "Hey database developers, a little help!"

Later . . . Jim

Java Search

I think I'm finally getting the hang of Object Oriented programming. I have been working on the user interface to provide all administration from it and add quite a bit of usability.

Over the past few days, I added search to the playback window. Because the playback window has two frames (Text classes in Java), this is a bit trickier than your standard text search. To begin with, I am allowing the search to be limited to one or the other frame, as well as using both. This means that the search class has to be aware of each frame.

To be able to highlight the text, the Text class actually needs to be defined as StyledText. When the text is found, the replaceStyleRanges method is called to highlight it. For now, I am leaving it highlighted, thinking that it is more helpful to be able to see all of the found selections. The current found text is displayed in reverse video by using the setSelection method. This one has to be reset by setting the selection range to zero before setting the new selection.

I thought about being able to use a single find window to search multiple playback windows, but this made my head hurt. However, it did seem friendly to share the search strings between playback windows. So I created a string array in the global variables class, and store the strings there. I even save them to the preference store so that they are maintained over application restarts.

If you are interested in this code, check out the WindowPlayback*.java source modules in the subversion repository. The preference store is defined and initialized in the Globals.java and MainWindow.java modules.

The beauty of the OOP style is that almost all of my code is spent managing indexes. The heavy lifting is done by other classes and their methods. So I am hopeful that these GUI enhancements will be finished pretty soon, and I will build another set of download packages by the end of the month.

Later . . . Jim

Whatcha Doin'?

As soon as I put the latest download on SourceForge, I started working on the user interface. I am hoping to bring it up to a version 1.0 level of usability. At the rate I am going, I expect to have a download ready in about a month.

So that's what's coming. But what I barely mentioned, was a new feature in the last package that I added a week or so before it was built. The reason was, I wanted to see how well it worked in the pilot project before explaining it. But now I have done that, and I am pretty happy with it.

The feature is a comparison of the client and server session size. When a TCP session ends, the client data is multiplied by a factor, and if that is larger than the server data, it is reported. The multipliers are specified by port, so each protocol can be handled uniquely.

In the pilot, I have set port 80 to have a multiplier of 1, just as a proof of concept. There have not been too many reports on it, and those can be grouped in two categories:

• Document uploads: This was what I was hoping to see. And I believe that it means if there is an exploit that loads a rootkit, that would be reported, as well.


• Large cookies: I never realized how much is saved in some cookies. I have seen several (Facebook, I'm looking at you!) that are over 4Kbytes.


• Miscellaneous: The rest of the reports are usually client requests for many files, but some of them don't get sent for some reason.

Again, this is mostly a proof of concept feature and I hope to expand on it down the road. But it gives me the sense that the Realeyes IDS is capable of detecting behavior. I think that's pretty cool.

Later . . . Jim

New Realeyes IDS Release and Demos

There is a new download available for Realeyes v0.9.4. Read the release notes for details, but basically there are several fixes and a few new features.

Unfortunately, while testing the packages, I discovered that the previous release had a ridiculous error that caused the system to fail if only IPv4 or IPv6 were chosen to be monitored. My lame excuse for this is that I made some significant changes to the database and spent a lot of time testing those. When I modified the install script to allow for alternate combinations of IPv4 and IPv6, I 'assumed' that the original case would work.

Another problem I discovered is that between the last release (Sept. 2008) and now, the eclipse project archived the version of the SWT libraries I was pointing to from the installation instructions and the Microsoft Windows installer. So that has been corrected, and hopefully won't change again.

Essentially it comes down to people and their (my) limitations. I was planning to add a page to the project web site to encourage developers, but I made it a priority after finding that embarrassing glitch.

I am very pleased to announce that I have created several demos of the application. The main links go to youtube, but there are ogg theora versions available also. And, if anything, these are better than those at youtube. Assuming you have the mplayer plugin installed in your browser, you should get a video that fills the window, which makes it much easier to see than the little viewer on the youtube page.

The demos were created using the GTK version of recordMyDesktop. This is a simple application, but it does what it is supposed to very well. The only technical issue I had was that the output file name needs to be set before the recording.

The biggest problem I had was in my performance. I wrote outlines for the demos, but discovered that even just moving the cursor around, much less opening and closing windows, caused me to lose track of what I was saying. So I eventually wrote out scripts which I nearly read word-for-word. And even at that, there are noticeable pauses in a couple of places. If possible, I recommend that 2 people work together, one talking, the other manipulating the screen.

Anyhow, I hope these demos give people an idea of the power of this system. I am finding quite a bit that is interesting at the college where I am running a pilot project. I will be writing about that soon.

Later . . . Jim

Punishment vs. Prevention

Punishment

Recently, F-Secure released a report titled, "Growth in Internet crime calls for growth in punishment". The article and the associated report cite F-Secure's research and several specific incidents to make the case for creating an 'Internetpol' to fight cybercrime. It is their conclusion that "against a background of steeply increasing Internet crime, the obvious inefficiency of the international and national authorities in catching, prosecuting and sentencing Internet criminals is a problem that needs to be solved."

The data that is used to reach this conclusion is tenuous at best. The primary fact cited is that the number of signatures used in the F-Secure detection database has increased three times over a year ago. This could be explained in many ways, with one of the main ones being that exploit creators have adapted to signature based detection by automatically generating variations of the original, which requires many more signatures to detect a single basic exploit. Numbers alone do not tell the story.

As a side note, this is one of the strengths of the Realeyes IDS. While the rules include specific characters to be matched, they can be detected in any order and then correlated with an Action. At the next level, multiple Actions can be correlated with an Event. This allows many variations of an exploit to be defined by a single Event rule.

F-Secure's anecdotal evidence of outbreaks is even less convincing. It is just as easy to conclude that attacks are more targetted than a few years ago, when a single worm could infect millions of systems, and infer from this that software development has become at least good enough to deter the easy attacks. Yet, neither scenario is absolutely supported by the evidence.

But even if the problem were defined correctly, the solution presented is not. First and foremost, what is a cybercrime in international terms? Most countries have not updated their own laws to meet the conditions presented by the Internet. The thought that Brazil, Russia, India, China, the UK, the US, and all the other countries with Internet access could agree on a common set of laws to govern Internet usage is a stretch, to say the least.

Then there is the issue of prosecution. The situation of a perpetrator who is in a different country from the computers attacked would probably not be any different from how that is handled today. And it is all too common for bureaucratic agencies to use quantity instead of quality to prove success. This initiative would very likely result in many low-level 'criminals' and even some innocent people being swept up in the new dragnet.

Finally, I find it extremely simplistic to suggest dumping society's problems on law enforcement. A huge question is how this internetpol organization would be staffed, especially considering that existing law enforcement agencies are finding it challenging to enforce existing laws in the Internet environment. Between jurisdictional issues and competition for the qualified candidates, the new agency would certainly create inefficiencies. And where is the funding to come from?

Prevention

I believe that legislatures need to update laws to define what is cyber crime. The recent case on cyberbullying has produced potentially bad precedents that need to be addressed, and soon. But most of this effort should focus on adapting current law to the Internet and only creating new laws where they are justified by a unique situation.

The truth is, much of the problem is technological. SQL injection attacks are an example. Currently, every application programmer is expected to parse input for this. But many application programmers hardly know what a database is, much less how to protect against all the possible variations of SQL injection. The ones who do know that are the database developers. Therefore, the security community should be calling for all xDBC libraries to include methods to validate input for applications.

The F-Secure report cited botnets as one of the primary security concerns. The root cause of botnets is spam Email. If this were not such a lucrative business, it would not be such a problem. One of the solutions is to force strong authentication in Email protocols. And this is just one example. The security community should support an organization that could act as consultants to protocol committees to define strong security solutions for Internet protocols. That organization could also focus on convincing vendors and users to implement those solutions.

There are many guides on secure programming, but how many application developers have studied them? This should be mandatory, because if exploiting vulnerabilities were hard, there would not be nearly as many attacks. The security community could help produce more secure applications by establishing a certification program for secure programming.

Realistically however, the biggest part of the problem is unaware users. We in the industry talk about best practices, but that is meaningless to most users. We need to convince management to ensure that users get adequate training about good security practices and we need to be specific about what that training includes.

Finally, I feel compelled to issue the warning, "Be careful what you wish for, because you might just get it." If the government takes over Internet security, there is sure to be a large amount of new regulation imposed. And this could mean security companies like F-Secure would have to devote a lot of resources towards compliance. I think it would be much better for us to take responsibility for finding solutions ourselves.

Later . . . Jim

More Results from Realeyes

For the past few weeks, I have been learning a lot about the site where the Realeyes pilot project is being run. After seeing several reports of incidents from Europe and Asia, it occurred to me that I could create a rule to monitor non-US IP addresses.

To do this, I got the IANA IPv4 Address Assignments, and created a list of the high order octet assigned to each of Europe, Asia, Africa, and Latin America. The rule was simply a SYN packet and any match on the first octet of the source address with a value in the list.

At first, I simply turned it loose, which generated over 20,000 reports. I was able to reduce that quickly by filtering on the "Referer:" field. First were the requests being referred by one of the site's own web servers. Then I found other sites, such as Google, that were referring browsers to the monitored network. These were all defined in a single Action , which was then defined in the Event with the 'NOT' flag set. This resulted in about 5,000 reports which have been further reduced by adding some of the site's commonly requested web pages to the filter.

The rule is now: Any connection requested by an IP address in Europe, Asia, Africa, or Latin America that is not referred by a site server or one in a list of other servers, and is not requesting one of a list of web pages. If a match is found, the first 64k of both halves of the session are reported. I was thinking of adding a filter where the web server responds with a 200 message, but that could miss a successful exploit.

Of the reported sessions, many are web crawlers for various international search engines. A large number are being referred by other servers. And a fair number appear to be overseas students. Many of the web crawler and overseas student connections consisted of over 100 sessions. Using the 'Ignore Source Address' option, I could close the incidents for a single source IP address without creating a report in a single click. This allowed me to reduce the reports by 2,000 - 3,000 fairly quickly.

And that left me with about 1,000 connections of 1 - 5 sessions each. It was easy to display the playback window and see the client request and the server response and make a decision on whether it was valid or not. Usually, the server responded with a 200 code and sent the requested page. I was able to check about 10 of these per minute, so it only took a couple of hours to run through the entire list.

As far as invalid activity, there have been several targetted scans. By this I mean that the requests are only sent to web servers, and they actually make HTTP requests. These were easy to see by sorting the reports on the source IP address and looking for connections to multiple servers.

The most interesting one was 'GET /manager/html'. This appears to be a Tomcat exploit which tries to gain access to the administrator account. Of the dozen web servers that received this request, all but one replied with 404 "Not Found". The other one replied with 401 "Unauthorized" and the source host then sent over 150 variations of the authorization code field. The codes were mixtures of numbers and mixed case letters that looked like they were taken from a table. Some were as long as 25 characters, while others were only 5 or 6 characters. Fortunately, none were successful.

Another interesting discovery was that one of the monitored site's web servers was being used to store data. An application to allow students to participate in workgroup activities had been broken into and data was stored for some of those sites that are links in spam. It was the response from the server that alerted me to this. I saw a list of keywords meant to generate a lot of hits in search engines. I was then able to report the full path of the request to the web administrator and the server was cleaned of this and a few other pages.

The lesson I take from this is that Realeyes is capable of collecting a broad range of data, filtering it effectively, and providing enough information to analysts to very quickly determine the severity level of incidents. The rules for monitoring can be customized and tuned for the site's requirements, giving analysts and administrators a deep view of their environment. And since that is what I set out to do with this project, I am quite pleased.

Later . . . Jim

It's a Big Cloud

I have recently read several articles that comment on the issues surrounding 'cloud computing'. However, they all seem to be the proverbial blind men describing an elephant. Doc Searls covers more ground than most and promises a follow up discussion, but all of them tend to limit the issues to their own perspective.

I don't have a problem with their facts, just the level of incompleteness. First, I'd like a really good definition of 'cloud computing'. Since I have not seen one, I will take a shot at it. As a network guy, I have seen clouds in network diagrams for a long time, so I tend to build on that understanding to relate to the current state of the technology.

The essence of 'cloud computing' is that it extends the personal computer to utilize Internet resources. These days, a personal computer may be a desktop, a kiosk station, a laptop, a mini-laptop, a smart phone, or an applicance. The resources that can be used include services, such as weather or stock information, interactive applications, such as Email or social network sites, storage sites, such as Flickr, or computer-to-computer applications, such as tracking packages or vehicles.

Using this definition, it is obvious that 'cloud computing' is simply a buzzword. Anyone who makes online purchases, has an online Email account, or has joined a social networking site is participating in 'cloud computing'. Even if the requirement that data that would otherwise be stored on a local disk be involved, all three of these examples meet the definition.

So what's the big deal? Well, as usual, social mores are behind the technology, and some of the discussion is about trying to catch up. Also, if the definition can be controlled, it can be sold as a new product, with the inescapable "caveat emptor" warnings from consumer advocates. With that in mind, I see the following issues involved in using this technology. Not surprisingly, many of them are the same as the issues of using computer technology in general, but the addition of the Internet puts a new spin on them:

• Cost: The proponents of 'cloud computing' tend to tout this as a big plus. That sounds to me like they are trying to sell a web version of out-sourcing. The easiest argument against it would be that traditional out-sourcing has not proven to be a huge cost saver across the board.
  
  But the real issue is the question of, "Who is the target market?" I cannot imagine a retail company putting it's inventory on storage managed by Google. So realistically, the target market is individual consumers. The type of data being handled is mostly Email, audio/video files, and blogs. For this, the online storage -- and backups -- are very cost effective.
  
  Will we ever see companies out-sourcing to web services/storage? I never say never, but I think it is a really hard sell. So I predict that it will eventually take hold, but in a limited way. Applications that help companies interact with their customers could be beneficial to both parties. And then there are the ones that no one has thought of yet.


• Reliability: Adding the Internet to the equation makes
  reliability a huge issue. The components that must all be working are:
  
  
  
  • The personal computer
    
  • The local ISP
    
  • The cloud
    
  • The remote ISP
    
  • The remote services
    
  
  The further out you go, the higher the possibility of failure. But so what? How many times in your life have you missed a critical phone call or Email, where minutes or even hours made a difference? In my mind, the backups at the storage site are far superior to the procedures done by the majority of consumers, myself included. And this outweighs the few times that the site is inaccessible, and is more cost effective at the same time. Even downtime for businesses would not lead to their demise.
  


• Access to data: The remaining items are where much of the current discussion is centered. There are some online Email services where it is difficult to retrieve Emails to the personal computer. But this is an issue that can be managed. It essentially boils down to read before you sign, and if you're the type who doesn't do this, well shame on you. Also, it would be pretty silly to not keep a copy of at least the most important data locally, such as photos, which adds to the cost. But the cost of losing it forever is even higher.
  


• Privacy: If there is anyone who hasn't figured it out yet, let me put it as plainly as possible. Nothing on the Internet is private. All the privacy policies and laws in the world cannot stop someone who has a real desire to take whatever data they want and do whatever they want with it. Data that is more important, such as financial information can be more carefully protected (this is where I get to plug Realeyes), but computer security is a matter of probabilities, not guarantees. As far as what the Internet companies do with your data, again, read before you sign. But if there is information that should never be exposed, it should never be accessible from the Internet.
  


• Ownership: I believe that this is the reason that Richard Stallman said using 'cloud computing' is stupid. He has long championed freedom of information, not just program source code. The ownership policies of most sites promoting 'cloud computing' goes against this in that they consider your information to be free, but not their own. Therefore, his position is consistent and reasonable. However, if you consider losing $1,000 in Las Vegas to be part of the fun, then I guess you can be forgiven for thinking he is a spoil sport.
  
  It is true that most of the sites that handle consumer data reserve the right to use that data as they please. Their literature basically says this is for promotional purposes. And Google finds keywords in Email to display ads. I have known people who take the dealer logo off their cars because they object to being a billboard. If that is you, you are probably going to be uncomfortable using these sites. But having a social networking account is not a right. You have to pay to play. Just remember the rule about Internet privacy, which is that there is none.
  


• Security: This is not a rehash of the storage site's security policies. It is a serving of food for thought. The Internet is a virtual wild west. The quality of security from one site to the next varies widely, and there are many who are happy to take advantage of that. The more you use the Internet, the higher your chances of having a vulnerability exploited. So if you use the same personal computer for social networking that you use for banking, you may become a victim of identity theft. I can think of a couple of analogies here, and I expect that you can too. I would say that the same types of rules apply. At the very least, please keep your security patches up to date.
  

To 'cloud compute' or not, that is the question. Of course, if you are reading this, you have already answered it. Now it is simply a matter of degree. Are you going to participate fully, or limit your involvement to a few interests? The most important thing is to be aware of the issues. I hope that I have contributed some useful thoughts to that end.

Later . . . Jim

Realeyes in the Real World

For about a year, I have been running Realeyes in a pilot project at a local college. The first six months were spent making it scale to a real world environment. But for the past six months, I have been able to experiment with rules to achieve the objectives of the system:

• Capture exploits: This one is obvious, but that doesn't make it easy.


• Reduce false positives: This was an original design goal, and my testing shows a lot of promise.


• Determine if exploits are successful: This was also an original design goal, and it too is showing promise.


• Capture zero-day exploits: While this capability hasn't been designed into the system, the flexibility of the rules and the statistics collection are showing promise in accomplishing this. We have also found some incidents of ongoing invalid activity that was apparently flying under the rader.

There is not much I can say about writing rules to detect exploits. If you have never done it, you just can't imagine how hard it is. First, finding enough information to even start is very difficult. But even with decent information, picking out the critical pieces that will consistently reveal the exploit is much more of an art than a science. I have written a few dozen rules at this point, but certainly don't consider myself to be an expert. With that said, rules do get written and, after some trial and error, are fairly effective.

The holy grail of creating IDS rules is to increase the capture rate while simultaneously reducing the false positive rate. When I was on the front lines, monitoring the IDSes, there was always the fear that a major attack would be buried in the false positives. Therefore, when I started this project, my personal goal was to completely eliminate false positives. I may not succeed, but I figure that if I am shooting for 100%, I will get closer than if my target is a lot lower.

I had spent quite a bit of time thinking about and discussing this with others. The team I worked with used multiple IDSes and correlated the information from them. Our success rate was very good, compared to our counterparts in other agencies. So a part of the puzzle seemed to be collecting enough data to see the incident for what it really is.

As an analogy, imagine taking a sheet of paper and punching a hole in it with a pencil. Then try to read a magazine article through that hole. At the very least, I would want to make the hole bigger. But this is as far as the analogy takes me, because what I really need is context, and a single word or phrase doesn't do that for me.

But even using multiple IDSes wasn't the complete solution. Each IDS had limitations that reduced its contributions to the total picture. If a rule was producing too many false positives, then the rule had to be modified, often in a way that reduced its effectiveness. This meant that there were important pieces missing.

So the solution appeared to be, put the capabilities of all the IDSes in a single IDS and do the correlation at the point of data capture. And that is how the Realeyes Analysis Engine is designed. Three levels of data that are analyzed: Triggers, Actions, and Events. Each one feeds the next, and the correlation of multiple inputs gives a broader view of the activity.

Triggers are the smallest unit of data detected. They can be header data such as a source or destination port, a string such as "admin.php" or "\xeb\x03\x59\xeb\x05\xe8\xf8", or a special condition such as an invalid TCP option, handled by special code. These are correlated by Action definitions, where an Action definition may be the Triggers: 'Dest Port 80' AND 'Admin PHP'.

Actions are the equivalent of rules in most IDSes available today. It is the next level of correlation that is showing the promise of reaching the goals listed above. Events are the correlation of Actions, possibly in both halves of a TCP session. (Actually, Realeyes assumes that UDP has sessions defined by the same 4-tuple as TCP.) And this is where is gets interesting.

One of the first rules defined was to search for FTP brute force login attempts. This Event is defined by two Actions, the first being 'Dest Port 21' and 'FTP Password', the second being 'Source Port 21' and 'FTP login errors' (which is any of message 332, 530, or 532). The rule was defined to require more than three login attempts to allow for typos. It has reported a few dozen incidents and zero false positives. Considering that the string for the FTP password is "PASS ", I am quite pleased with this result.

A more recent rule was defined to detect an SQL Injection exploit. The exploit uses the 'cast' function to convert hexadecimal values to ASCII text. The rule's Triggers search for four instances of testing for specific values in a table column. The Action is defined to fire if any two of them is found in any order. Although the exploit is sent by a browser to a server, there is no port defined. This rule is reporting a couple hundred attempts per day and none of them are false positives.

It really got exciting when the rule captured a server sending the exploit. It turned out that the server was simply echoing what it received from the browser in certain instances. However, the web admins wanted to know what web pages this was happening for, and this is where the third design goal is demonstrated. Both halves of these sessions were reported and displayed in the playback window. This gave us the full URL data being sent to the server, and the web admins were able to address the problem quickly.

To address the issue of detecting new exploits, Realeyes has a couple of somewhat unique features. The first is statistics collection. The information maintained for each session includes the number of bytes sent and received. When a session ends, these are added to the totals for the server port. Then, three times a day, the totals are reported to the database. This allows for the busiest ports to be displayed. Or the least busy, which might point to a back door.

But there are other statistics that can be collected, as well. It is possible to monitor a specific host or port and collect the total data for each session.

For example, I monitored UDP port 161, which is used for SNMP, and saw two host addresses using it that belonged to the site, but 3 or 4 a day that did not. Since there was not much traffic for the port, I simply created a rule to capture all data for it. This showed me that there were unauthorized attempts to read management data from devices in the network, but that none were successful.

Using the same technique, I monitored TCP port 22, used for SSH, and found several sessions that appeared to be attempts to steal private keys. I reported this to the admin of the target hosts, and while he had applied the most recent patches, I also suggested that he regenerate his private keys, to be on the safe side.

Another feature for discovering new exploits is time monitoring. This is setting a time range for all activity to a host or subnet to be captured. I defined rules to monitor the EMail servers between midnight and 5:00 am. The original intent was to watch for large bursts of outbound EMail to detect any site hosts being used for spam. We have found one of these.

But we have discovered several other things from this. First was a large amount of traffic using a port that the network admin thought had been discontinued. Second, there have been several attempts to have EMail forwarded through the site's servers. This may be a misconfiguration at the sender, or it may be an attempt to send spam through several hops to avoid detection. From this I created rules to monitor for the most common domains.

So far, there have not been any earth shattering discoveries (to my disappointment, but to the the admins' relief). But as I said, the signs are promising that the system is capable of meeting the design goals. Until a couple of weeks ago, I have spent the majority of my time working on code. But I am starting to spend more time on rules. So I am looking forward to making some new discoveries. Stay tuned.

Later . . . Jim

My App Fails the LSB

My position on the Linux Standard Base has evolved. When I first heard about it, I was all for it. The LSB as a standard could be useful to some, but now I disagree with the goals of the LSB working group. To be sure, this post is not about dissing the Linux Foundation. They have many worthwhile projects.

What follows is my experience with the LSB Application Checker, my take on the purpose of the LSB, and my own suggested solution for installing applications on GNU/Linux distributions. The Realeyes application failed to certify using the checker v2.0.3, which certifies against the LSB v3.2. Everything that it called out could be changed to pass the tests, but I will only consider correcting a few of the 'errors'.

After building the Realeyes v0.9.3 release, I collected all executable files in a common directory tree, downloaded the LSB application checker, and untarred it. The instructions say to run the Perl script, app-checker-start.pl, and a browser window should open. The browser window did not open, but a message was issued saying that I should connect to http://myhost:8889. This did work, and I was presented with the Application Check screen.

There was a text box to enter my application name for messages and one to enter the files to be tested. Fortunately, there was a button to select the files, and when I clicked on it a window opened that let me browse my file system to find the directories where the files were located. For each file to be tested, I clicked on the checkbox next to it, and was able to select all of the files, even though they were not all in the same directory. Then I clicked on the Finish button and all 87 of the selected files were displayed in the file list window.

When I clicked on the Run Test button, a list of about a dozen tasks was displayed. Each was highlighted as the test progressed. This took less than a minute. Then the results were displayed.

There were four tabs on the results page:

• Distribution Compatability: There were 27 GNU/Linux distributions checked, including 2 versions of Debian, 4 of Ubuntu, 3 of openSUSE, 3 of Fedora, etc. Realeyes passed with warnings on 14 and failed on the rest.


• Required Libraries: These are the external libraries required by the programs written in C. There were nine for Realeyes, and three (libcrypto, libssl, and libpcap) are not allowed by the LSB. This means that distros are not required to include the libs in a basic install, so they are not guaranteed to be available.


• Required interfaces: These are the calls to functions in the libraries. There were almost a thousand in all, and the interfaces in the libraries not allowed by the LSB were called out.


• LSB Certification: This is the meat of the report and is described in some detail below.

The test summary gives an overview of the issues:

• Incorrect program loader: Failures = 11


• Non-LSB library used: Failures = 4


• Non-LSB interface used: Failures = 60


• Bashism used in shell script: Failures = 21


• Non-LSB command used: Failures = 53


• Parse error: Failures = 5


• Other: Failures = 53

The C executables were built on a Debian Etch system and use /lib/ld-linux.so.2 instead of /lib/ld-lsb.so.3. The Non-LSB libraries and interfaces were described above, but there was an additional one. The Bashisms were all a case of either using the 'source' built in command or using a test like:

  while (( $PORT < 0 )) || (( 65535 < $PORT )); do

which, in other Bourne shells, requires a '$(( ... ))'. The parse errors were from using the OR ("||") symbol.

The fixes for these are:

• Use the recommended loader


• Statically link the Non-LSB libraries


• Use '.' instead of 'source'


• Rework the numeric test and OR condition

So far, all of this is doable, sort of. But every time a statically linked library is updated, the app must be rebuilt and updates sent out. Also, the additional non-LSB library is used by another one. So I would have to build that library myself and statically link the non-LSB library (which happens to be part of Xorg) in it (and that one is part of the GTK). The reason I am using it is because the user interface is built on the Eclipse SWT classes, which uses the local graphics system calls to build widgets.

The non-LSB commands include several Debian specific commands (such as adduser), and for my source packages I had to rework the scripts to allow for alternatives (such as useradd). But the other disallowed commands are:

• free: To display the amount of free memory


• sysctl: To set system values based on the available memory


• scp: Apparently the whole SSL issue is a can of worms


• java


• psql

Finally, all of the other 'errors' are: Failed to determine the file type, for the file types:

• JAR


• AWK


• SQL


• DTD and XML

Part of the problem with the LSB is that it has bitten off more than it can chew. Apparently Java apps are non-LSB compliant. So are apps written in PHP, Ruby, Erlang, Lisp, BASIC, Smalltalk, Tcl, Forth, REXX, S-Lang, Prolog, Awk, ... From my reading, Perl and Python are the only non-compiled languages that are supported by the LSB, but I don't know what that really means (although I heartily recommend to the Application Checker developers that they test all of the executables in their own app ;-). And I suspect that apps written in certain compiled languages, such as Pascal or Haskell, will run into many non-LSB library issues.

Then there are databases. Realeyes uses PostgreSQL, and provides scripts to build and maintain the schema. Because of changes at version 8, some of these scripts (roles defining table authorizations) only work for version 8.0+. The LSB Application Checker cannot give me a guarantee that these will work on all supported distros because it didn't test them. I have heard that there is some consideration being given to MySQL, but from what I can tell, it is only to certifying MySQL, not scripts to build a schema in MySQL.

After all this kvetching, I have to say that the Application Checker application is very well written. It works pretty much as advertised, it is fairly intuitive, and it provides enough information to resolve the issues reported by tests. My question is, "Why is so much effort being put into this when almost no one is using it?"

An argument can be made that the LSB helps keep the distros from becoming too different from each other and without the promise of certified apps, the distros would not be motivated to become compliant. But I only see about a dozen distros on the list, with Debian being noticeably absent. And yet, there is no more sign of fragmentation in the GNU/Linux world than there ever was.

My theory on why UNIX fragmented is that proprietary licenses prevented the sharing of information which led to major differences in the libraries, in spite of POSIX and other efforts to provide a common framework. In the GNU/Linux world, what reduces fragmentation is the GPL and other FOSS licenses, not the LSB. All distros are using most of the same libraries, and the differences in versions are not nearly as significant as every UNIX having libraries written from scratch.

I have to confess, I couldn't care less whether Realeyes is LSB compliant, because it is licensed under the GPL. Any distro that would like to package it is welcome. In fact, I will help them. That resolves all of the dependency issues.

While I am not a conspiracy theorist, I do believe in the law of unintended consequences. And I have a nagging feeling that the LSB could actually be detrimental to GNU/Linux. The only apps that benefit from LSB compliance are proprietary apps. The theory behind being LSB compliant is that proprietary apps can be guaranteed a successful installation on any LSB compliant GNU/Linux distro. I'm not arguing against proprietary apps. If a company can successfully sell them for GNU/Linux distros, more power to them. However, what if proprietary libraries manage to sneak in? This is where the biggest threat of fragmentation comes from.

But even more importantly, one of the most wonderful features of GNU/Linux distros is updates, especially security updates. They are all available from the same source, using the same package manager, with automatic notifications. If the LSB is successful, the result is an end run around package managers, and users get to deal with updates in the Balkanized way of other operating systems. That is a step in the wrong direction.

The right direction is to embrace and support the existing distro ecosystems. There should be a way for application teams to package their own apps for multiple distros, with repositories for all participating distros. The packages would be supported by the application development team, but would be as straightforward to install and update as distro supported packages.

There is such a utility, developed by the folks who created CUPS. It is called the ESP Package Manager. It claims to create packages for AIX, Debian GNU/Linux, FreeBSD, HP-UX, IRIX, Mac OS X, NetBSD, OpenBSD, Red Hat Linux, Slackware Linux, Solaris, and Tru64 UNIX. If the effort that has gone into LSB certification were put into this project or one like it, applications could be packaged for dozens of distros.

And these would not just be proprietary apps. There are many FOSS apps that don't get packaged by distros for various reasons, and they could be more widely distributed. Since the distros would get more apps without having to devote resources to building packages, they should be motivated to at least cooperate with the project. And don't forget the notification and availability of updates.

As a developer and longtime user of GNU/Linux ('95), I believe that all of the attempts to create a universal installer for GNU/Linux distros are misguided and should be discouraged. I say to developers, users, and the LSB working group, "Please use the package managers. A lot of effort has been put into making them the best at what they do."

Later . . . Jim

What is happening to IPv6?

I have read several articles lately on the state of IPv6. Most of them take the position that implementation is moving slowly, and that is just fine. I went to a conference a few weeks ago that gave me a fairly different impression. It brought together people from the US federal government, business, and academia to describe the state of IPv6 in the government, and start a discussion of where to go next. My main interest in the conference was the security issues surrounding IPv6, and of course, there is no serious discussion about the Internet that doesn't include security. However, I did learn some interesting things about the status of IPv6.


At the time, the government was on the verge of announcing that agencies could demonstrate IPv6 capabilty. This basically amounts to the fact that they have a plan in place to only purchase equipment that supports IPv6, although some have pilot programs in place. The DoD is apparently furthest along, but is nowhere near ready to "flip the switch". However, there has been quite a bit of planning and NIST is working on setting standards for testing equipment, that will eventually lead to purchase specifications.

The government, and in particular the DoD, is not very concerned about the increased address space or encryption/authentication features of IPv6. Instead, their interest is in mobile networks, especially the auto-configuration features, and multicast. This is for using dynamic networks on the battlefield, and while they didn't give any details on the subject, it isn't hard to imagine how both features could be used. There is a pretty good article from Linux Journal on IPv6 mobility in Linux, which starts with a really good overview of the principles of using it.

Near the end of the conference, there was a panel discussion which included a couple of government representatives, a guy from Cisco and one from AT&T, and a member of the IETF IPv6 working group. The guys from industry said that there is a steady, and fairly significant increase in the IPv6 network infrastructure. If you live in the US, you are probably used to reading articles that claim deployments are almost non-existant and are surprised to hear this. However, the deployments are happening in Asia primarily, and to a lesser extent in Europe.

I don't know how true it is, but someone at the conference said that the University of Califonia system has more IPv4 addresses than the entire country of China. Even if it isn't the case, the point is well taken. Latecomers to the Internet got the crumbs of the IPv4 address space, and now they are desperate to use IPv6.

This has many implications. The entire set of TCP/IP and related protocols are very loose. In some people's minds, the rollout of IPv6 is an opportunity to get several things right that IPv4 didn't. However, implementations will not all support the same features. The reason for this is that IPv6 has a common header with very little in it other than the source and destination addresses. The new features are in extension headers, which are optional. And the protocols to be used for encryption and authentication are continuing to be developed:

• Security Architecture for IP: RFC1825 obsoleted by RFC2401 obsoleted by RFC4301 (IPSec) and updated by RFC3168


• IP Authentication Header: RFC1826 obsoleted by RFC2402 obsoleted by RFC4302 and RFC4305 obsoleted by RFC4835

The newest of these is dated April, 2007. And none of them specify algorithms or methods, just the requirements for using them. Meanwhile, thousands of routers for IPv6 networks have already been deployed and the number is growing steadily.

The question was raised during the panel discussion of how much assurance agencies will have that the products they purchase will support the requirements that are mandated. The answer was a resounding, "Well ..." The member of the IETF pointed out that they are not a standards committee, but more like the maintainer of the documentation. RFCs have their status upgraded to Standard by concensus. And the Asian and other countries using IPv6 as 'IPv4 with big addresses' are not interested in spending time haggling over features that interfere with their deployments.

But as I said before, my interest was in the security concerns for IPv6. The main ones are the same as in IPv4. To begin with, a very small percentage of exploits go against the IP layer. And security issues for the upper layers of the stack will not change.

But once mobile IPv6 networks with auto-configuration begin to be deployed, there will be vulnerabilities to be dealt with. Of course, the first vulnerability will be admins on the learning curve. Hopefully, these folks will be motivated and supported to get good training early on. To get a taste of actual protocol issues, read the RFC for Secure Neighbor Discovery (SEND) mechanisms.

I certainly hope that the admins don't get stuck with the bulk of the responsibility for security. But the current state of networking products, according to someone who is testing them for the NSA, is that the low level infrastructure (routers, host OSes) is fairly good, but not 100%. Even a little way up the stack, that falls to little, if any, support for IPv6. This includes firewalls and IDSes, and some server applications.

So what has to be handled? Well, the first step is to find the upper level data, such as a TCP header. This means parsing the extension headers, which is a little weird. First, the header type is not in each header, but in the preceeding header. So the first two octets of each extension header are the type of the next header and the length of the current header.

Then there is tunneling. It is possible to tunnel IPv6 in IPv4 or vice versa. What this means for finding the upper level header is that first the IPvX header must be parsed, then the IPvY header. And this could be done more than once in crafted packets. I sure hope that normal traffic at the point of a firewall would not have multiple IPv4 and IPv6 tunnels.

The other issue is the order and number of IPv6 extension headers. Most packets are only allowed to have one of each type of extension header. The exception is the Destinations header, and there may be two of these. Oh, and the order is not mandatory, just recommended.

And what is the status of Realeyes IPv6 support? I have added the capability to find upper level headers in untunneled packets. And there is some capability to parse IPv6 extension headers. In particular, the number of each type of header is validated, and for now the recommended order is treated as mandatory, so a report is issued for packets that don't follow the recommendation.

The common IPv6 header is formatted in playbacks, but there is a hex dump of the extension header data. I will be gradually adding support for tunneling and formatting of the extension header data. It is similar to TCP options, so I don't see the code being difficult. The biggest problem is testing. I configured two hosts on my local network to talk to each other with IPv6, but that is a proof of concept, at best.

So that is my perspective of what is happening with IPv6. I suggest that if you work on networks, you show more than a passing interest in it, because it will probably be a case of hurry up and wait, until all of a sudden we need it yesterday.

Later . . . Jim

Realeyes IDS v0.9.3 Ready for Download

The packages for Realeyes IDS release 0.9.3 are ready for download. This version improves stability over the previous one and includes several new features.

If there are any problems, please notify the team. Thanks.

You may have noticed that I finally finished the First Edition of the Realeyes IDS Manual. I made it as useful as I could. But any comments are welcome.

Later . . . Jim

Building a Debian GNU/Linux package

I recently built packages for the 0.9.3 release of Realeyes. These include both source and Debian GNU/Linux packages. For reasons that I have forgotten, I built the Debian packages before I started working on the source packages, but I'm glad that I did. Going through that process made me add several things that I would have overlooked, especially man pages.

I actually built an entire distribution as part of the process of creating my packages. It does not meet the requirements for re-distribution, but is very handy for laying down a fresh install with exactly what is needed to run Realeyes. I will provide the steps for that in another post.

The source is comparatively easy, once all of the files are collected, a tar file is created of the directory. The trick here is to verify that the C code will compile. I have read a lot of comments about how straight-forward the standard configure/make installation procedure is. But from a developer's perspective, there are several issues, mainly having to do with autoconf and automake. I don't have enough space here to discuss these, and I really don't have any wisdom to impart if I did, because I have only learned as much as I needed for my own packages.

Debian packaging is somewhat more complicated. A package that is included in a Debian distribution must go through fairly rigorous tests. There are several scripts for checking correct packaging procedures, including lintian and linda. These verify such conditions as:

• executables are built correctly


• man pages exist for every executable file


• naming conventions are followed


• Debian documentation is formatted correctly

I have seen a few HOWTOs on building a Debian package, and there is a huge amount of information on the Debian site. But I still had to piece together a working plan for myself, so I thought it would be worth sharing it. There may be better ways to do some of it, but I have written scripts that get me through the process with only a little manual effort. And even when there is an automated process, it is still good to know what is happening under the hood. So without further ado, I offer my experience.

Building a Debian Package

I. Read enough of the manuals to get a sense how Debian packages are built, and then keep links to them for reference:

• The Debian Reference manual


• The Debian Policy manual


• The Debian Developers reference

II. Create a working directory for the package

• Get a package to use as a model, using the following commands to extract the package files and the Debian metadata files:
  

  
  apt-get -d -y --reinstall install package_name
  dpkg -x package_name 
  cd package_dir
  dpkg -e ../package_name



• Create a working directory, and under it make the directories to be installed for the package, even if there are no files to be saved in them. These may include:
  
  working_dir/etc/package_name
  working_dir/etc/init.d
  working_dir/usr/sbin
  working_dir/usr/share/package_name
  working_dir/usr/share/doc/package_name
  working_dir/usr/share/man
  working_dir/var/log/package_name


• Make the Debian control directory with its files (as needed)
  
  working_dir/DEBIAN
  
  
  • control: This contains the description of the package, including dependencies, architecture, and the description used by aptitude or synaptic -- use the model to create this for the first time
    
  
  
  • conffiles: This contains any configuration files installed with the package -- I put mine in /etc/realeyes
    
  
  
  • preinst: This is a shell script that runs before the package is installed, if it exists
    
  
  
  • postinst: This is a shell script that runs after the package is installed, if it exists -- I use it to create user IDs
    
  
  
  • prerm: This is a shell script that runs before the package is de-installed, if it exists
  
  
  • postrm: This is a shell script that runs after the package is de-installed, if it exists


• Populate the directories with the application files: Use the model to help understand what goes where
  

III. Use the maintainer tools to verify package acceptability

• Build the package:
  

  cd working_dir
  dpkg-deb --build package_dir package_name
  

• lintian/linda: Check for package discrepancies. Note that lintian and linda are not in the standard package and must be installed separately. Lintian uses Perl and linda use Python, so there may be several dependencies installed with them.
  

  lintian -i package_name > package_name.lintian
  linda package_name > package_name.linda
  

• Fix all the problems, here are some helpful hints from my experience:
  
  
  • Man pages: txt2man is a program that takes ascii text and converts it to a man page. It works for simple pages, but the resulting groff file may have to be edited manually in some cases. Use 'gzip -9' to compress man pages.
  
  
  • Compiled programs: Compiled programs must be stripped. Use the command:
    

    install -s
    

  • Identify all non-executable files in system directories (ie. /etc/package_name) in the package's DEBIAN/conffiles.
  
  
  • Lintian provides the section in the Debian Policy manual that describes the requirement that was flagged.
    


• Sign the package: Create a GPG key for the package and sign each package with the key
  

  gpg --gen-key
  dpkg-sig -s builder pkg.deb
  

  NOTES:
  
  
  • The public keyring is in $HOME/.gnupg/pubring.gpg
  
  
  • There should be a lot of entropy on the system to help the random number generator, (grep -R abc /usr/* seems to work well)
  
  
  • Issue the command 'cat /proc/sys/kernel/random/entropy_avail' to find out how much entropy is currently available, it should be at least above 1,500
    

At this point, the package can be installed using the dpkg command. However, if there are dependencies that must be installed, dpkg will issue a warning about them, but does not handle their installation. So if you want to go to the next level, here is what you have to do.

IV. Repository directories

A Debian repository has a relatively simple directory structure to maintain the packages and metadata about them. An installation ISO is basically a repository tree with just the stable packages. A custom repository tree can be added to the apt sources.list to be accessed just like officially maintained packages, with aptitude or synaptic.

In the top repository directory, the following are mandatory:

• md5sum.txt: The list of all files in the tree with their md5 checksums
  
  To create the md5sum.txt file for my mini-distro, I wrote a script that ran in the top ISO directory. It did a recursive ls, ran md5sum on all regular files, and wrote the output to the md5sum.txt file. I keep that as a template and only update the files that change.
  


• pool: The subdirectory where packages are kept. Under the pool directory, there are a few pre-defined directories where the different categories of packages are kept. Anyone who has edited a sources.list file has seen most of these:
  
  
  • main: Technically, these are packages that meet the Debian Free Software Guidelines (DFSG), but I think of them as the officially maintained packages
  
  
  • contrib: Contributed packages are DFSG, but depend on packages that are not -- I use this for my own packages, even though I don't have any non-DFSG dependencies
    
  
  
  • non-free: These are non-DFSG packages
  
  The structure of each of these is the same. The package directories are in subdirectories named with the first letter of the package name. The exception is libraries, which are under directories named libletter, which is the prefix of the library package name. Below these subdirectories are the directories with the actual package files.


• dists: The dists directory contains the metadata about packages. There is a directory for the distribution, in this case, etch. In an ISO, there are also the directories, frozen, stable, testing, and unstable, which are links to the distribution directory. In a repository, these may have their own files. But for my purposes, I only include the distribution subdirectory.
  
  Under the distribution subdirectory are the following:
  
  
  • Release: This file describes the packages, including the architecture, the components, and contains md5sums for the package metadata files -- the file information also includes the file size, and since there are only a few of these, I created the original by hand
    
  
  
  • main: This directory contains the metadata about the main packages
  
  
  • contrib: This directory contains the metadata about the contrib packages
  
  The structure of main and contrib is the same, and again, I only use contrib. The architecture directories are below contrib, and in my case, there is only binary-i386. In the architecture directory there are three files:
  
  
  • Packages: This uses information from the DEBIAN/control file and adds such things as the full path of the package file
  
  
  • Packages.gz
  
  
  • Release: This contains metadata about the contrib directory


• I also put a few optional files in the top level directory. These include a copy of the GPL (I use version 3), installation instructions, and the public key for the signed packages. The installation instructions explain how to add the public key so that aptitude and synaptic can validate the packages.
  

V. Add the packages

• Copy the packages to the appropriate pool directory. In my case, this means copying them to:
  
  
  iso_dir/pool/contrib/r/realeyescomponent


• Create an override file for the packages. This consists of a line for each packages with the following information:
  
  
  package priority section
  
  In my case it looks like this:
  

  realeyesDB   optional  net
  realeyesDBD  optional  net
  realeyesGUI  optional  net
  realeyesIDS  optional  net
  

  The man page for dpkg-scanpackages (the next command) has a description of each field and says that the override file for official packages is in the indices directory on Debian mirrors.
  


• Build the metadata:
  

  dpkg-scanpackages \
  pool/contrib/ override.etch.contrib > \
  dists/etch/contrib/binary-i386/Packages
  cd dists/etch/contrib/binary-i386
  gzip -c -9 Packages > Packages.gz
  cd ../..
  md5sum contrib/binary-i386/Packages* > md5.tmp
  ls -l contrib/binary-i386/Packages* >> md5.tmp
  

• The file, md5.tmp, is edited to put the file size after the md5 checksum and before the file name, and the file listing lines are deleted. Then the file, Release, is edited to read in the file, md5.tmp, at the end and the duplicate lines are deleted.
  

  gpg --sign -ba -o Release.gpg Release
  cd ../..
  cp -a ~/.gnupg/pubring.gpg RE_pubring.gpg
  md5sum ./INSTALL* > md5sum.txt
  md5sum ./GPL* >> md5sum.txt
  md5sum ./dists/etch/Release >> md5sum.txt
  md5sum ./dists/etch/contrib/binary-i386/* >> md5sum.txt
  md5sum ./pool/contrib/r/*/* >> md5sum.txt

VI. Installing the package

The instructions for installing this package are:

• Copy the debian packages to a directory that will be used for the initial installation and future updates, such as, /var/tmp/realeyes. Untar the packages:
  

  tar xvzf realeyes_debian.tar.gz
  

• Change to the top level packages directory and add the public key file, RE_pubring.gpg, to the Debian trusted sources with the command:
  

  apt-key add RE_pubring.gpg
  

• Edit the file /etc/apt/sources.list to add the line:
  

  deb file://install_dir/realeyes/ etch contrib
  

• Update the package lists with one of the following methods:
  
  
  • On the command line enter:
    

    apt-get update
    

  • In aptitude, select Actions -> Update package list


• Install the package using aptitude or synaptic

So there you have it. I hope it shortcuts the learning curve a little.

Later . . . Jim

Security Unobscured

I just read this post about the state of information security. There are two main points that seem to be related, and a third that is implied.

First, Jason feels that security professionals are not being taken seriously by other IT professionals. Maybe it's because I live in the Washington, DC area, but I don't have that impression. However, if it is true at all, I can't help but think that it is because of some people in the field shouting, "The sky is falling," combined with an ever increasing number of real security breaches that are poorly addressed, ranging from web based exploits and phishing scams to stolen laptops containing the unencrypted personal information of thousands of people. It makes the general public wonder, "What do those security people do?"

I had a discussion with an IT professional the other day about a recent SQL injection exploit. Over time, the lower levels of the stack, including the operating system and server daemons, have become fairly hard targets, while applications are still pretty soft. And predictably, the attacks are moving up the stack. A large part of the problem is that many (if not most) programmers have little understanding of the infrastructure on which their application is built, in this case, databases. But it is left to the programmers to test the data for SQL injection attempts. Why isn't the security community clamoring for database APIs to include a higher degree of data assurance, as well as more security in other APIs?

And that segues into Jason's second point, "the market is flooded with these so called CISSP certified IS professionals". I get the impression that he considers them to be charlatans. And in the era of terrorist attacks and Sarbanes-Oxley, there are IT professionals and managers who are wary of buying IS snake oil. Referring back to my comment about database APIs, the IT security field must be seen to be solving security problems. Instead, vendors and consultants often seem to be peddling their own wares, while the security of people's information is a secondary concern. In fact for some, the increase of real threats almost seems to be a marketing tool.

And that brings me to the third point that Jason implied when he said, "I remember a day when security people were feared." Having been on both sides of this fence, I see a tension between IT security and performance, with performance almost always being the top priority. In my experience, the security people don't get their own budget, they get the leftovers from the system and network budget -- the exception being financial institutions and some government agencies. After all, security is a lot like insurance, and everyone prefers the cheapest insurance, especially when the economy is down.

While I agree with Jason that IT professionals should take the security professional's concerns seriously, I think it is even more important for security people to focus on solutions. Most IT people look at security breaches the same way they look at hardware failures, a problem to be solved as quickly as possible, so that the real work of their environment can continue. The security people will get more respect by helping to keep these to a minimum than by coming on as tyrannical hall monitors.

We all know that there is no security silver bullet, it requires that the people using and maintaining the systems do the right thing. But if they don't know what the right thing is, because of poor training, being unaware of potential hazards, or lack of information about their environment, the security professionals have to take some responsibility for it. Computer technology is still a relatively young field, and both innovation and expansion are changing the landscape every year. The job of security professionals will continue to be providing security solutions for current conditions by doing research, explaining the findings, and developing tools to help the people in the trenches.

My own approach is to give the IT people as much useful information as possible. The reason I started Realeyes was to reduce false positives and provide the admins with enough data to make quick decisions and increase awareness of what is happening in their networks. They know the systems and networks best, so it is my job to give them the best tools to do their job right.

If security people want attaboys, we have to provide more light and less shadows.


Later . . . Jim