about summary refs log tree commit diff stats
path: root/difference.cpp
diff options
context:
space:
mode:
authorKelly Rauchenberger <fefferburbia@gmail.com>2016-04-16 14:22:25 -0400
committerKelly Rauchenberger <fefferburbia@gmail.com>2016-04-16 14:22:25 -0400
commitff6d29e7f6b587a2536227834950986dbbcd580b (patch)
tree11baa1d8a29ccbc6bdfd7a1323361ab28a26d499 /difference.cpp
parentf7b91944738e732ab4bfea50ea0a2fffd92a51a6 (diff)
downloaddifference-ff6d29e7f6b587a2536227834950986dbbcd580b.tar.gz
difference-ff6d29e7f6b587a2536227834950986dbbcd580b.tar.bz2
difference-ff6d29e7f6b587a2536227834950986dbbcd580b.zip
Added Accept header to image requests
The canonical bot tweeted an image (https://twitter.com/differencebot/status/721395886291558400) containing an advertisement instead of the requisite object. Previously, the only defense against servers serving the wrong image was that we ignore 300 response codes. This image, when loaded in Google Chrome, loaded a document with a content type of text/html, which is also ignored by difference, and which executed JavaScript redirecting Chrome to a malware-infested page. difference, however, saw the response as an image with content type image/gif (notably different from the URL, which indicated a JPEG image). It turned out that Chrome was using an Accept header that prioritized text/html documents over most other content types, which the malicious server used to decide what content to serve. Changing difference to send the same header caused the malicious server to also serve the text/html document to difference, which difference then discarded. Whilst the Accept header being used now does prioritize text/html documents over images, servers with legitimate content will not use that information when deciding what document to serve.

The malicious test URL is http://www.northvalleymedicalsupply.com/shop/products_pictures/adj%20hinge%20knee%20brace.jpg.
Diffstat (limited to 'difference.cpp')
-rw-r--r--difference.cpp9
1 files changed, 9 insertions, 0 deletions
diff --git a/difference.cpp b/difference.cpp index 66e4550..7ea8b74 100644 --- a/difference.cpp +++ b/difference.cpp
@@ -94,6 +94,11 @@ int main(int argc, char** argv)
94 } 94 }
95 } 95 }
96 96
97 // Accept string from Google Chrome
98 std::string accept = "Accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8";
99 curl::curl_header headers;
100 headers.add(accept);
101
97 std::cout << "Started!" << std::endl; 102 std::cout << "Started!" << std::endl;
98 for (;;) 103 for (;;)
99 { 104 {
@@ -153,6 +158,8 @@ int main(int argc, char** argv)
153 curl::curl_ios<std::ostringstream> img1ios(img1buf); 158 curl::curl_ios<std::ostringstream> img1ios(img1buf);
154 curl::curl_easy img1handle(img1ios); 159 curl::curl_easy img1handle(img1ios);
155 std::string img1url = lstvec[curind]; 160 std::string img1url = lstvec[curind];
161
162 img1handle.add<CURLOPT_HTTPHEADER>(headers.get());
156 img1handle.add<CURLOPT_URL>(img1url.c_str()); 163 img1handle.add<CURLOPT_URL>(img1url.c_str());
157 img1handle.add<CURLOPT_CONNECTTIMEOUT>(30); 164 img1handle.add<CURLOPT_CONNECTTIMEOUT>(30);
158 165
@@ -202,6 +209,8 @@ int main(int argc, char** argv)
202 curl::curl_ios<std::ostringstream> img2ios(img2buf); 209 curl::curl_ios<std::ostringstream> img2ios(img2buf);
203 curl::curl_easy img2handle(img2ios); 210 curl::curl_easy img2handle(img2ios);
204 std::string img2url = lstvec[curind]; 211 std::string img2url = lstvec[curind];
212
213 img2handle.add<CURLOPT_HTTPHEADER>(headers.get());
205 img2handle.add<CURLOPT_URL>(img2url.c_str()); 214 img2handle.add<CURLOPT_URL>(img2url.c_str());
206 img2handle.add<CURLOPT_CONNECTTIMEOUT>(30); 215 img2handle.add<CURLOPT_CONNECTTIMEOUT>(30);
207 216