PolarProxy 1.0 Released

I am thrilled to announce the release of PolarProxy version 1.0 today! Several bugs that affected performance, stability and memory usage have now been resolved in our TLS inspection proxy. PolarProxy has also been updated with better logic for importing external root CA certificates and the HAProxy implementation has been improved. But the most significant addition in the 1.0 release is what we call the “TLS Firewall” mode.

TLS Firewall

PolarProxy now supports rule based logic for determining if a session should be allowed to pass through, get blocked or if the TLS encrypted data should be inspected (i.e. decrypted and re-encrypted) by the proxy. This rule based logic can be used to turn PolarProxy into a TLS firewall. As an example, the ruleset-block-malicious.json ruleset included in the new PolarProxy release blocks traffic to malicious domains in abuse.ch’s ThreatFox IOC database as well as traffic to web tracker domains listed in the EasyPrivacy filter from EasyList. This ruleset also includes an allow list in order to avoid accidentally blocking access to legitimate websites.

PolarProxy’s ruleset logic isn’t limited to just domain names. It is also possible to match traffic based on JA3 or JA4 hashes as well as application layer protocol information provided in the ALPN extension of a client’s TLS handshake.

For more information on the ruleset format and how to use PolarProxy as a TLS firewall, see here:
https://www.netresec.com/?page=TlsFirewall

Linux, macOS and Windows builds of the new PolarProxy release can be downloaded from here:
https://www.netresec.com/?page=PolarProxy

Posted by Erik Hjelmvik on Thursday, 02 May 2024 07:00:00 (UTC/GMT)

Tags: #PolarProxy​ #TLS​ #inspect​ #bypass​ #ThreatFox​ #ASCII-art​

TLS Redirection and Dynamic Decryption Bypass in PolarProxy

PolarProxy is constantly being updated with new features, enhanced performance and bug fixes, but these updates are not always communicated other than as a short mention in the ChangeLog. I would therefore like to highlight a few recent additions to PolarProxy in this blog post.

Custom TLS Redirection

One new feature in PolarProxy is the --redirect argument, which can be used to redirect TLS traffic destined for a specific domain name to a different domain. This feature can be used to redirect TLS-encrypted malware traffic going to a known C2 domain to a local HTTPS sandbox instead, for example INetSim.

This --redirect argument will cause PolarProxy to terminate outgoing TLS traffic to malware-c2.com and redirect the decrypted traffic into a new TLS session going to inetsim.local instead. The “--leafcert noclone” argument forces PolarProxy to generate a fake X.509 certificate for “malware-c2.com” rather than sending a clone of the certificate received from the INetSim server to the malware implant.

Note: You also need to specify a proxy mode, such as -p for transparent proxy or --socks for SOCKS proxy, to make the command above work.

The --redirect argument can also be used to perform domain fronting, which is a clever method for hiding the true destination of HTTPS based communication, in order to circumvent censorship or for other reasons conceal who you’re communicating with. The following command can be used to set up a local socks proxy that redirects traffic destined for YouTube to google.com instead:

A browser configured to use PolarProxy as a SOCKS proxy will send HTTPS requests for youtube.com to PolarProxy, which then decrypts the TLS layer and re-encrypts the HTTP communication in a new TLS session directed at google.com instead. Someone who monitors the outgoing traffic from PolarProxy will assume that this is normal Google traffic, since the SNI as well as certificate will be for google.com. On the server side however, after having decrypted the TLS layer, Google will kindly forward the client’s original HTTP request for youtube.com to an endpoint that serves the content for YouTube.

Dynamic TLS Decryption Bypass

PolarProxy is designed to block TLS connections that it can’t decrypt, except for when the server’s domain name is explicitly marked for decryption bypass with the “--bypass” command line argument. However, as of recently PolarProxy also supports dynamic TLS decryption bypass using a form of fail-open mode. When this fail-open mode is enabled, PolarProxy attempts to intercept and decrypt proxied TLS traffic, but allows connections to bypass decryption if the same client-server pair has previously rejected PolarProxy’s certificate. This method is convenient when monitoring network traffic from applications that enforce certificate pinning or for some other reason can’t be configured to trust PolarProxy’s root CA – provided that it’s acceptable to let traffic that can’t be decrypted to pass through untouched rather than blocking it, of course.

The following command line option configures PolarProxy to allow new TLS connections to bypass decryption for one hour (3600 seconds) after previously having failed to decrypt traffic between the same client and server.

A simple way to verify this fail-open feature is to do a simple test with curl. It doesn’t matter if the client you’re testing on is Windows, Linux or macOS, since PolarProxy as well as curl is available for all three platforms.

Web browsers that don’t trust PolarProxy’s root CA will display a certificate warning the first time they visit a website that PolarProxy tries to decrypt traffic for.

But once the dynamic bypass has kicked in the user will no longer see a certificate warning when visiting the same website again, since traffic between that client and server is now end-to-end encrypted.

Handling of non-TLS traffic and Better Logging

Other new features in PolarProxy is the “--nontls” argument, which can be used to specify how to handle connections that doesn’t use TLS. The default action is to block non-TLS connections, but they can also be allowed to pass through (if the target host is known) or to forward the connection to a specific host and port. There is even a “--nontls encrypt” argument, which can be used to encrypt traffic that isn’t already TLS-encrypted before forwarding it to a specific host. This feature can be used as an alternative to stunnel to wrap traffic from applications that lack TLS support inside a TLS tunnel.

PolarProxy now also produces less output to stdout, unless -v is used, and error messages have been improved to be more specific and easier to understand.

Posted by Erik Hjelmvik on Tuesday, 28 February 2023 13:42:00 (UTC/GMT)

Tags: #PolarProxy​ #TLS​ #redirect​ #bypass​ #SNI​ #ASCII-art​

PolarProxy 0.8.16 Released

Client Certificates

PolarProxy now supports client-authenticated TLS handshakes for outgoing connections to support sites that require mutual TLS (mTLS) authentication. The following example uses the PKCS#12 client certificate "client.p12" with password "pwd" to authenticate PolarProxy when connecting to "https://api.example.com":

Thanks to Peter Lambrechtsen for the idea!

Bypassing Decryption for Specific Domains

There are situations when it isn't appropriate to decrypt the traffic passing through PolarProxy. The traffic might, for example, contain personal or confidential information. It might also not be possible to decrypt the traffic for technical reasons, such as when clients use certificate pinning or certificate transparency to validate the server certificate. We therefore recommend that such sites are put on a "bypass" list, i.e. a list of domains for which PolarProxy should let the encrypted traffic pass untouched to preserve the end-to-end encryption between the client and server.

PolarProxy's "--bypass <file>" option, which can be used to provide a regular expression list of domains not to decrypt, has now been acompanied by "--bypassexact <file>". The new --bypassexact option simply matches domains against the lines in <file> using string matching of the full domain name, no fancy-pants regex involved.

PCAP-over-IP Client

The new "--pcapoveripconnect" option can be used to let PolarProxy connect to a PCAP-over-IP listener and send it a live PCAP stream of decrypted traffic over TCP. This option complements PolarProxy's "--pcapoverip" option, which sets up a PCAP-over-IP listener that serves clients with the same PCAP stream. Thanks to Andy Wick for suggesting adding a PCAP-over-IP connector to PolarProxy!

The following command instructs PolarProxy to send a live PCAP stream with decrypted traffic to a local PCAP-over-IP listener:

PolarProxy will automatically attempt to re-establish the PCAP-over-IP connection every 10 seconds if it goes down or cannot be established for some reason.

Only Store Packets When Instructed

PolarProxy no longer writes hourly rotated pcap files with decrypted packets to disk unless explicitly instructed to do so with "-o <directory>" or "-w <file>".

Flushing Buffered Packets to Disk

PolarProxy now periodically flushes buffered packets to disk every 60 seconds. The flush interval can be controlled with the "--autoflush <seconds>" option. The auto flush can also be disabled with "--autoflush 0".

No More Out-of-Quota Issues

We have also improved the quota handling for our privileged users, who have a license key that allows them to decrypt more than 10 GB or 10 000 TLS sessions per day. You should now be able to use your full daily quota without issues!

UPDATE 2022-12-08

Peter Lambrechtsen's talk IoT your Pet from Kawaiicon 2022 is on YouTube! In this talk Peter explains how he used PolarProxy to MITM traffic between an IoT device and a cloud service running on Amazon AWS. Check out Peter's Pet Hub Local project for more details.

Posted by Erik Hjelmvik on Monday, 30 November 2020 07:45:00 (UTC/GMT)

Tags: #Netresec​ #PolarProxy​ #PCAP​ #TLS​ #bypass​ #PCAP-over-IP​ #pcapoverip​ #certificate​