The four vulnerabilities discovered in Telegrams bespoke MTProto protocol highlight the dangers of trying to invent new cryptographic systems rather than using proven existing solutions. MTProto is used by Telegram as a replacement to the industry standard TLS protocol.
Four vulnerabilities were discovered during the research and have now been addressed by Telegram in app updates (7.8.1 for Android, 7.8.3 for iOS and 2.8.8 for Telegram Desktop). They included:
1, An attacker could re-order messages coming from a client to the server without detection.
This allows, for example, to alter the order of “pizza” and “crime” in the sequence of messages: “I say yes to”, “all the pizzas”, “I say no to”, “all the crimes”. This attack is trivial to carry out.
~ Martin R. Albrecht (Information Security Group, Royal Holloway, University of London)
2, An attacker could detect whether a message acknowledgement was sent and received
3, It is theoretically possible (but not really practically possible) for an attacker to recover some of the plaintext from an encrypted message due to the complex nature of the way the protocol is implemented and the order which certain steps must be completed. Proving once more that complexity is the enemy of security.
4, It is theoretically possible to mount a man-in-the-middle attack on the initial key negotiation between the client and server and then impersonate the server to the messaging client breaking the confidentiality and integrity of the communications.
The write up of the attacks performed by the researchers serves to highlight that even if a cryptographic function is secure, it can still be implemented in such a way as to leak information. Which just adds to the risk when creating home grown crypto as you will necessarily be doing things for the first time and making mistakes along the way – the best practice has not yet been invented for your new system. Home grown crypto systems are a red flag to wise Security Managers and should be approached with caution and avoided in most business situation.
Telegram has addressed all the issues identified during the research in the latest versions of their apps, and has posted an explanation and rebuttal of the findings they did not agree with.
Remi Vaughn from Telegram said:
The traits of MTProto pointed out by the group of researchers from the University of London and ETH Zurich were not critical, as they didn’t allow anyone to decipher Telegram messages. That said, we welcome any research that helps make our protocol even more
These particular findings helped further improve the theoretical security of the protocol: the latest versions of official Telegram apps already contain the changes that make the four observations made by the researchers no longer relevant.