The dl module in Ruby 1.8.5 and earlier, 1.8.6 through 1.8.6-p286, 1.8.7 through 1.8.7-p71, and 1.9 through r18423 does not check "taintness" of inputs, which allows context-dependent attackers to bypass safe levels and execute dangerous functions by accessing a library using DL.dlopen.

Ruby 1.8.5 and earlier, 1.8.6 through 1.8.6-p286, 1.8.7 through 1.8.7-p71, and 1.9 through r18423 does not properly restrict access to critical variables and methods at various safe levels, which allows context-dependent attackers to bypass intended access restrictions via (1) untracevar, (2) $PROGRAMNAME, and (3) syslog at safe level 4, and (4) insecure methods at safe levels 1 through 3.

Algorithmic complexity vulnerability in the WEBrick::HTTPUtils.splitheadervalue function in WEBrick::HTTP::DefaultFileHandler in WEBrick in Ruby 1.8.5 and earlier, 1.8.6 through 1.8.6-p286, 1.8.7 through 1.8.7-p71, and 1.9 through r18423 allows context-dependent attackers to cause a denial of service (CPU consumption) via a crafted HTTP request that is processed by a backtracking regular expression.

The regular expression engine (regex.c) in Ruby 1.8.5 and earlier, 1.8.6 through 1.8.6-p286, 1.8.7 through 1.8.7-p71, and 1.9 through r18423 allows remote attackers to cause a denial of service (infinite loop and crash) via multiple long requests to a Ruby socket, related to memory allocation failure, and as demonstrated against Webrick.

Integer overflow in the (1) rbarysplice function in Ruby 1.8.4 and earlier, 1.8.5 before 1.8.5-p231, 1.8.6 before 1.8.6-p230, 1.8.7 before 1.8.7-p22, and 1.9.0 before 1.9.0-2; and (2) the rbaryreplace function in 1.6.x allows context-dependent attackers to trigger memory corruption, aka the "beg + rlen" issue. NOTE: as of 20080624, there has been inconsistent usage of multiple CVE identifiers related to Ruby. The CVE description should be regarded as authoritative, although it is likely to change.

The rbstrformat function in Ruby 1.8.4 and earlier, 1.8.5 before 1.8.5-p231, 1.8.6 before 1.8.6-p230, 1.8.7 before 1.8.7-p22, and 1.9.0 before 1.9.0-2 allows context-dependent attackers to trigger memory corruption via unspecified vectors related to alloca, a different issue than CVE-2008-2662, CVE-2008-2663, and CVE-2008-2725. NOTE: as of 20080624, there has been inconsistent usage of multiple CVE identifiers related to Ruby. The CVE description should be regarded as authoritative, although it is likely to change.

Multiple integer overflows in the rbstrbufappend function in Ruby 1.8.4 and earlier, 1.8.5 before 1.8.5-p231, 1.8.6 before 1.8.6-p230, 1.8.7 before 1.8.7-p22, and 1.9.0 before 1.9.0-2 allow context-dependent attackers to execute arbitrary code or cause a denial of service via unknown vectors that trigger memory corruption, a different issue than CVE-2008-2663, CVE-2008-2664, and CVE-2008-2725. NOTE: as of 20080624, there has been inconsistent usage of multiple CVE identifiers related to Ruby. This CVE description should be regarded as authoritative, although it is likely to change.

Directory traversal vulnerability in WEBrick in Ruby 1.8.4 and earlier, 1.8.5 before 1.8.5-p231, 1.8.6 before 1.8.6-p230, 1.8.7 before 1.8.7-p22, and 1.9.0 before 1.9.0-2, when using NTFS or FAT filesystems, allows remote attackers to read arbitrary CGI files via a trailing (1) + (plus), (2) %2b (encoded plus), (3) . (dot), (4) %2e (encoded dot), or (5) %20 (encoded space) character in the URI, possibly related to the WEBrick::HTTPServlet::FileHandler and WEBrick::HTTPServer.new functionality and the :DocumentRoot option.

Directory traversal vulnerability in WEBrick in Ruby 1.8 before 1.8.5-p115 and 1.8.6-p114, and 1.9 through 1.9.0-1, when running on systems that support backslash (\) path separators or case-insensitive file names, allows remote attackers to access arbitrary files via (1) "..%5c" (encoded backslash) sequences or (2) filenames that match patterns in the :NondisclosureName option.

The SecureRandom.randombytes function in lib/securerandom.rb in Ruby before 1.8.7-p352 and 1.9.x before 1.9.2-p290 relies on PID values for initialization, which makes it easier for context-dependent attackers to predict the result string by leveraging knowledge of random strings obtained in an earlier process with the same PID.

Drew Yao of Apple Product Security reported a flaw in how Ruby's BigDecimal implementation. In a 64-bit program, if the amount of memory to allocate is greater than 2^32, the allocated amount will be truncated to a 32-bit unsigned integer, however the amount to memset will use the full 64-bit number, leading to a crash or memory corruption. After the memset, the unexpectedly small buffer could be used in other ways that lead to memory corruption or, possibly, code execution.

This flaw only affects 64-bit processes.

To test whether or not ruby is vulnerable, the following can be used (on a vulnerable ruby it will cause a segfault):

$ ruby -e 'require "bigdecimal"; BigDecimal.new("8").(0x20000000)'

Upstream committed the following to fix the flaw:

http://svn.ruby-lang.org/cgi-bin/viewvc.cgi/trunk/ext/bigdecimal/bigdecimal.c?r1=29364&r2=30993