This article is more than 1 year old
TLS proxies: Insecure by design, say boffins
Home antivirus is rubbish, so you don't use it on your work PCs. Do you?
Have you ever suspected filters that decrypt traffic of being insecure? Canadian boffins agree with you, and have said that TLS proxies – commonly deployed in both business and home networks for traffic inspection – open up cans of worms.
“Not a single TLS proxy implementation is secure with respect to all of our tests, sometimes leading to trivial server impersonation under an active man-in-the-middle attack, as soon as the product is installed on a system," wrote Xavier de Carné de Carnavalet and Mohammad Mannan of the Concordia Institute of Systems Engineering in Montreal.
The trio's paper (PDF) goes on to say that users could be exposed to man-in-the-middle attacks or other CA-based impersonations.
While the researchers focused their attention on consumer anti-virus and Web filtering products, The Register reckons at least one of their warnings is borne out in the enterprise space.
It's not enough that vulnerabilities like POODLE and FREAK are patched by project maintainers, for example: customers of downstream products that happen to use the buggy software have to wait until their specific patches are available.
As de Carnavalet and Mannan write, outdated proxies might “lack support for safe protocol versions and cipher suites, undermining the significant effort spent on securing web browsers.“
The researchers tested eight antivirus and four parental control products (all for Windows), and two products that import a CA root certificate, checking their vulnerability for MITM attacks.
The depressing assertion:
We found that four products are vulnerable to full server impersonation under an active man-in-the-middle (MITM) attack out-of-the-box, and two more if TLS filtering is enabled. Several of these tools also mislead browsers into believing that a TLS connection is more secure than it actually is, by e.g., artificially upgrading a server’s TLS version at the client.
There's also the matter of how products protect their root certificates' private key. It's not pretty, as the table below shows.
Location | Protection | Access | |
---|---|---|---|
Avast | CAPI | Exportable key | Admin |
AVG | Config file | Obfuscation | Unknown |
BitDefender | DER file | Hardcoded passphrase | User |
BullGuard AV | DER reg key | Hardcoded passphrase | User |
BullGuard IS | DER reg key | Hardcoded passphrase | User |
CYBERsitter CER | file | Plaintext | User |
Dr Web | CAPI-cert1 | Exportable key | Admin |
ESET | CAPI | Non-exportable key | Admin |
G DATA | Registry | Obfuscated encryption | User |
Kaspersky | DER file | Plaintext | User |
KinderGate | CER file | Plaintext | User |
Net Nanny | Database | Modified SQL Cipher | User |
PC Pandora | CAPI-cert | Non-exportable key | Admin |
ZoneAlarm | DER file | Plaintext | User |
Although key recovery was non-trivial, the authors note “we retrieved four passphrase-protected private keys and a key stored in a custom encrypted SQLCipher database”.
Here's another treat: “The version of Kaspersky we analyzed in March 2015 continues to act as a TLS proxy when a 30-day trial period is expired; however, after the license expiration, it accepts all certificates, including the invalid ones”. Although fixed later, people who installed the pre-fix product and never uninstalled it are therefore MITM-vulnerability.
For the home user, de Carnavalet reckons filters that simply block domain names are probably effective enough.
The Register keenly hopes de Carnavalet and Mannan get the chance to repeat their tests against corporate proxies, and will keep some popcorn for just such an event. ®