New Ideas/Feature Requests

Hi Gary,

in regards to your requests.. #1 and #3 already exist .. #2 I'm not sure about that, however product mangers actively review features and consider all ideas posted.

Here is a brief tutorial on how I troubleshoot the SWA ..

In order to do this you will need to export the log file to a syslog server, I recommend a linux based system so you can easily use tools like "grep, cut, sed, awk etc"

Once your logs are uploaded you can see the entire format here:  http://swa.sophos.com/webhelp/swa/concepts/InterpretingLogFiles.html

now for the trouble shooting goodies:

the first question you need to ask is:

#1

is the site https or http? if https the first thing you can do is create an https scanning exemption, then a certificate exemption.. if the site immediately works you can verify the sites internet with tools like https://www.ssllabs.com/ssltest/index.html  test the target site and see if there are redirection issues, certificate issues or the site shows up with anything less than a C .. chances are the site is been blocked for cert issues.

#2

the next thing to try is to create a local site list entry and set it to "trusted" this will disable av scanning and allow clients to make "byte range requests" it will also disable AV scanning, in cases where the AV scanner hangs on streaming content.

#3

the next thing to try is setting the browser to use proxy.. set it to point directly to the appliance IP on port 8080.. if your using transparent redirection and explicit proxy works.. 99% chance there is some sort of redirection issue.  The most common issues is "the wccp service needs a quick restart" perhaps a service group is not working (granted this is more wide spreed then a simple web site) 

- private tabs will ensure content is not been cached and the equivalent of a 307 redirect is failing or serving bad data.  (also make sure caching is NOT enabled on the applaince its self in the advanced section)

also if clients are in full web control see this document:  https://community.sophos.com/kb/en-us/122384

ensure the workstation and your authentication and dns is as such: https://community.sophos.com/kb/en-us/126599 

check out the notes in my deployment guide: https://community.sophos.com/kb/en-us/126692

for the most part these are 1 off kb's and once your all set up you shouldn't need these anymore, other than reference.

If the above fails ..

now its time to check logs..

http://swa.sophos.com/webhelp/swa/concepts/InterpretingLogFiles.html

EP=1 SXL=1 h=10.99.115.13 u="DOMAIN\\johnsmith" s=200 X=- t=1336666489 T=284453
Ts=0 act=1 cat="0x220000002a" app="-" rsn=- threat="-" type="text/html" ctype="text/html"
sav-ev=4.77 sav-dv=2012.5.10.4770003 uri-dv=- cache=- in=1255 out=26198
meth=GET ref="-" ua="Mozilla/5.0 (Windows NT 6.1; WOW64; rv:12.0) Gecko/20100101 Firefox/12.0"
req="GET http://www.google.ca/ HTTP/1.1" dom="google.ca" filetype="-" rule="0"
filesize=25815 axtime=0.048193 fttime=0.049360 scantime=0.011 src_cat="0x2f0000002a"
labs_cat="0x2f0000002a" dcat_prox="-" target_ip="74.125.127.94" labs_rule_id="0"
reqtime=0.027 adtime=0.001625 ftbypass=- os=Windows authn=53 auth_by=portal_cache 
dnstime=0.000197 quotatime=- sandbox=-

EP/SXL = if you see this in the logfile the client is in full web control.. the appliance ignores ANY request that's tagged with endpoint as policy is already run on the request

h=10.99.115.13 u="DOMAIN\\johnsmith" shows the user that's cached and ip ... if there is no user "-" the ip is not authenticated and its probably hitting your default policy or blocking
s=200 this is the RCF internet code returned from the site ... 200 is good.. 300 is some sort of redirect .. 4/500 is generally bad.
t=1333666489 = the exact time in epoc .. you can use a converter or the date command to convert it right to the second.  On linux box you could sed/awk/sort the exact time you wish.

this is the first major part.. I would do something like : grep -i "[-7..0]" sophos_log  ..  this will sort out any request that was dropped either for policy or perhaps bad certs
etc.

it must be matched against the rsn= field.

the most common ones are 1406, blocked by policy or 1408 byte range requests and 1407

what I would normally do is..
live monitor the logs and then grep out the ip and sort by the epoc time.. 
what you will see is often something like

packet 1.. is request for url.xxx.com
packet 2.. is url xxx.com making a request to yyy.com
packet 3.. is act=-1 rsn 1406 for example.

this would be a typical request where say the inital page is NOT blocked.. but the site its self spiders out and pulls down content from a site that is blocked.

a good example:
if your watching a youtube video.. and the video never starts..
you could set it to trusted and make sure its not hanging from av scanning.. but in the back ground what is going on is.. you make a request for a video.. thats fine.. the video makes a back end request to an AD .
and thats blocked.. so the connection is never returned to the original request.. meaning the video never starts.

an added example would be something like, the back end server that is hosting the AD.. has a bad or self signed certificate.. so now the appliance blocks it because of that. 


the last thing we really care about int the log file is the category and risk class.

cat="0x220000002a" 
0x will always be the same
22 means low risk 21 would be trusted 23 would be high risk etc. 
2a is the actual category of the site

again when you looking at the set of packets.. you may see something like 
packet 1: 0x21000001a
packet 2: 0x22000001a
packet 3: 0x21000001a
packet 4: 0x24000001a

in this case you would want to white list the sites that came in as 22 .. (the change takes 5 mins to take effect)  get as many sites in the initial test as "trusted" to help troubleshoot
in the case of packet #4 .. its high risk.. risk class ALWAYS trumps policy.. anything this is globally blocked, or high risk will be blocked before policy is even run.

once the site works.. change the risk class back until you find the right one. 
generally this is not required.. but can be a very useful way to find the exact thing thats blocked.. could be a .jpg .. or a link.. or a back end server or anything else
along those lines.


Final notes:

keep in mind.. only the 5 browsers should ever go through the appliance.. things like fax machines or network printers or anything else that's not a user and joined
to the domain should be excluded from filtering.

so no applications, you can try making user agent string exclusions for authentication as well.

this should give you everything you need to troubleshoot any issue with the SWA..  


Unfortunately it's not exactly "easy" and sometimes opening a support case and watching the logs in real time is simply better.. 

Cheers

Hi Gary,

in regards to your requests.. #1 and #3 already exist .. #2 I'm not sure about that, however product mangers actively review features and consider all ideas posted.

Here is a brief tutorial on how I troubleshoot the SWA ..

In order to do this you will need to export the log file to a syslog server, I recommend a linux based system so you can easily use tools like "grep, cut, sed, awk etc"

Once your logs are uploaded you can see the entire format here:  http://swa.sophos.com/webhelp/swa/concepts/InterpretingLogFiles.html

now for the trouble shooting goodies:

the first question you need to ask is:

#1

is the site https or http? if https the first thing you can do is create an https scanning exemption, then a certificate exemption.. if the site immediately works you can verify the sites internet with tools like https://www.ssllabs.com/ssltest/index.html  test the target site and see if there are redirection issues, certificate issues or the site shows up with anything less than a C .. chances are the site is been blocked for cert issues.

#2

the next thing to try is to create a local site list entry and set it to "trusted" this will disable av scanning and allow clients to make "byte range requests" it will also disable AV scanning, in cases where the AV scanner hangs on streaming content.

#3

the next thing to try is setting the browser to use proxy.. set it to point directly to the appliance IP on port 8080.. if your using transparent redirection and explicit proxy works.. 99% chance there is some sort of redirection issue.  The most common issues is "the wccp service needs a quick restart" perhaps a service group is not working (granted this is more wide spreed then a simple web site) 

- private tabs will ensure content is not been cached and the equivalent of a 307 redirect is failing or serving bad data.  (also make sure caching is NOT enabled on the applaince its self in the advanced section)

also if clients are in full web control see this document:  https://community.sophos.com/kb/en-us/122384

ensure the workstation and your authentication and dns is as such: https://community.sophos.com/kb/en-us/126599 

check out the notes in my deployment guide: https://community.sophos.com/kb/en-us/126692

for the most part these are 1 off kb's and once your all set up you shouldn't need these anymore, other than reference.

If the above fails ..

now its time to check logs..

http://swa.sophos.com/webhelp/swa/concepts/InterpretingLogFiles.html

EP=1 SXL=1 h=10.99.115.13 u="DOMAIN\\johnsmith" s=200 X=- t=1336666489 T=284453
Ts=0 act=1 cat="0x220000002a" app="-" rsn=- threat="-" type="text/html" ctype="text/html"
sav-ev=4.77 sav-dv=2012.5.10.4770003 uri-dv=- cache=- in=1255 out=26198
meth=GET ref="-" ua="Mozilla/5.0 (Windows NT 6.1; WOW64; rv:12.0) Gecko/20100101 Firefox/12.0"
req="GET http://www.google.ca/ HTTP/1.1" dom="google.ca" filetype="-" rule="0"
filesize=25815 axtime=0.048193 fttime=0.049360 scantime=0.011 src_cat="0x2f0000002a"
labs_cat="0x2f0000002a" dcat_prox="-" target_ip="74.125.127.94" labs_rule_id="0"
reqtime=0.027 adtime=0.001625 ftbypass=- os=Windows authn=53 auth_by=portal_cache 
dnstime=0.000197 quotatime=- sandbox=-

EP/SXL = if you see this in the logfile the client is in full web control.. the appliance ignores ANY request that's tagged with endpoint as policy is already run on the request

h=10.99.115.13 u="DOMAIN\\johnsmith" shows the user that's cached and ip ... if there is no user "-" the ip is not authenticated and its probably hitting your default policy or blocking
s=200 this is the RCF internet code returned from the site ... 200 is good.. 300 is some sort of redirect .. 4/500 is generally bad.
t=1333666489 = the exact time in epoc .. you can use a converter or the date command to convert it right to the second.  On linux box you could sed/awk/sort the exact time you wish.

this is the first major part.. I would do something like : grep -i "[-7..0]" sophos_log  ..  this will sort out any request that was dropped either for policy or perhaps bad certs
etc.

it must be matched against the rsn= field.

the most common ones are 1406, blocked by policy or 1408 byte range requests and 1407

what I would normally do is..
live monitor the logs and then grep out the ip and sort by the epoc time.. 
what you will see is often something like

packet 1.. is request for url.xxx.com
packet 2.. is url xxx.com making a request to yyy.com
packet 3.. is act=-1 rsn 1406 for example.

this would be a typical request where say the inital page is NOT blocked.. but the site its self spiders out and pulls down content from a site that is blocked.

a good example:
if your watching a youtube video.. and the video never starts..
you could set it to trusted and make sure its not hanging from av scanning.. but in the back ground what is going on is.. you make a request for a video.. thats fine.. the video makes a back end request to an AD .
and thats blocked.. so the connection is never returned to the original request.. meaning the video never starts.

an added example would be something like, the back end server that is hosting the AD.. has a bad or self signed certificate.. so now the appliance blocks it because of that. 


the last thing we really care about int the log file is the category and risk class.

cat="0x220000002a" 
0x will always be the same
22 means low risk 21 would be trusted 23 would be high risk etc. 
2a is the actual category of the site

again when you looking at the set of packets.. you may see something like 
packet 1: 0x21000001a
packet 2: 0x22000001a
packet 3: 0x21000001a
packet 4: 0x24000001a

in this case you would want to white list the sites that came in as 22 .. (the change takes 5 mins to take effect)  get as many sites in the initial test as "trusted" to help troubleshoot
in the case of packet #4 .. its high risk.. risk class ALWAYS trumps policy.. anything this is globally blocked, or high risk will be blocked before policy is even run.

once the site works.. change the risk class back until you find the right one. 
generally this is not required.. but can be a very useful way to find the exact thing thats blocked.. could be a .jpg .. or a link.. or a back end server or anything else
along those lines.


Final notes:

keep in mind.. only the 5 browsers should ever go through the appliance.. things like fax machines or network printers or anything else that's not a user and joined
to the domain should be excluded from filtering.

so no applications, you can try making user agent string exclusions for authentication as well.

this should give you everything you need to troubleshoot any issue with the SWA..  


Unfortunately it's not exactly "easy" and sometimes opening a support case and watching the logs in real time is simply better.. 

Cheers

Hi Red_Warrior. I appreciate your response very much!

For number 1.

Ultimately what I am looking for is when I click on the search tab and search by “User” or “Site”, the dashboard will show me a status of Allowed or Blocked. There should be a separate column that shows me what triggered the blocked right there instead of having to go through the steps below. I am just too darn busy with all sorts of tasks to complete, not just managing and reviewing the SWA, so it’s important for me to be efficient as possible.

Regarding #2.

So many malware sites are newly registered domain names. Having a “Newly registered domain names” category would be huge in blocking malware and avoiding endpoint infections. Websense does this and the Sophos FW does this.

Regarding #3

Sometimes Endpoint A/V or another security device blocks malware distributing from a website while a user is browsing the Internet. The endpoint A/V or other security device will often just list the IP address and not know the website URL. One of the first things I want to know is which website is distributing malware so I can add an entry in the SWA local site list to block it. So I log on to Sophos SWA and search by user and narrow it down by the time that was previously listed by the security device to find out which website was distributing malware. The search results on the SWA may show 35 sites with the same time, such as 1:35 PM. How do I know which of these 35 sites was dishing out malware? Now if the search results would show milliseconds, such as 1:35:15 PM, like the security device, I would have it and could manually block the URL in the SWA.

Thanks again!

Gary

Hi Gary,

I think its important to note here, the Labs team monitors malware sources, spammers and has the very best access when it comes to malicious sites.

All of their work is packaged up in update packages and distributed to appliances and endpoint clients in near real time.

Tbh you really should not have to be constantly updating the local site list or trying to add domains manually. Just ensure there are no issues updating issues.

The effect of adding so many sites will be that the appliance will need to scan the enite local site list file for every request.. this may not be as efficient as you would like.. the other thing to note is you can only have 8192 entries..

To be totally honest, there is no reason you should ever need that many entries..  the vast majority of bad sites are collected from literally millions of submissions and updated via data updates every few mins.

All feature requests are always reviewed by developers and product managers.. I think you posted some good points..  So well see what happens

Cheers

1) Go to Search, by User, put in username and filter by status blocked.  The entries in the list should have status "Blocked (Virus)" and "Blocked (Policy)" etc.

2) Newly registered domains are not known by the categorizer appear as "Uncategorized".  Sophos collects the domain names of requests for all uncategorized sites and then attempts to categorize them correctly.  The SWA does not support a domain being multiple categories so we cannot support anything like "Alcohol and Newly Registered".  Therefore it is either new (uncategorized) or it is not new (categorized).

3) Every device has its own clock that can be off (even if using NTP) and the timestamp of detection can be different than the timestamp of downloading.  Timestamps are very useful to narrow down a general time window, but unless you are comparing timestamps from same device anything less than a second is useless.  Even down to the second is suspect.

swa t=1:10:20.300 User performed GET /malware.exe
swa t=1:10:24.500 SWA received response body from server and started malware scan
swa t=1:10:26.600 malware scan complete and clean, starting to send file to client
swa t=1:10:27.100 file completely sent and log made (using the 1:10:20.300 timestamp)

client T=1:10:26.300 file download complete in browser
client T=1:10:28.900 AV scanner on client detects virus
client T=1:10:30.100 AV scanner on client reports virus cleaned

So...  which timestamps are you going to match up?  The virus reporter on the client found something at 1:10:28.900 but the SWA log for the download is 1:10:20.300 (8 seconds off).  In order to keep requests from a client straight and clear as an "order of events" (as far as I remember) the SWA logs the request start time, not the response end time.  And even if it did, and if the clocks were sync'd perfectly, the AV detection time on the client could be seconds later.