ESSTCP: Enhanced Spread-Spectrum TCPAmir R. Khakpour, Hakima Chaouchi Institut National des Télécommunication (INT) Evry, France. Email:{amir.khakpour, hakima.chaouchi}@int-evry.fr Abstract Having stealth and lightweight authentication methods is empowering network administrators to shelter critical services from adversaries. Spread-Spectrum TCP (SSTCP) [1] is one of these methods by which the client sends an authentic sequence of SYN packets to the server for authentication. Since SSTCP have some certain drawbacks and security flaws, we propose an enhanced version of SSTCP (ESSTCP) which modifies the original algorithm to reduce the computational cost and cover its vulnerabilities from denial of service and replay attacks. Some performance problems like time synchronization are also resolved. We finally try to extend the functionality of this method for different applications and numbers of users by which ESSTCP can be performed as a secure Remote Procedure Call (RPC). 1. Introduction Hiding internet services from untrusted users would be one of the effective methods to protect not only the unpredictable attacks on local network and servers, but also to the unknown potential service and software vulnerabilities discovered gradually. Thereby, in order to distinguish between authorized users and adversaries, hidden authentication techniques should be exploited. These authentication techniques should be lightweight enough to be easily applicable on vast variety of devices and strong enough to be reliable for protecting crucial services and servers. Barham et al. [1] proposed a few techniques in which the client authenticates to firewall and asks for access to a specific port number for connection. Since this authentication has to be done stealthily, the general idea is to send some specific packets to closed ports of the firewall and trigger daemons on the firewall by authentic packets to open the desired ports for the authenticated user. These authentication techniques are classified into three groups based on the way the authentic packets are sent. In Spread-Spectrum TCP (SSTCP), the client sends a sequence of SYN packets to particular port numbers and prompts the firewall to execute the corresponding instructions for the client. In second method called Tailgate TCP (TGTCP), instead of sending a sequence of SYN packets, it sends a packet with some data including the secret and other parameters for authentication. In the third approach, Option-Keyed TCP (OKTCP), firewall only allows the SYN packets which contain a key encoded in some IP and TCP header fields to pass through. Subsequently, Port-knocking is introduced by Martin Krzywinski in [2] as a method of connection through a closed port. It is mostly focused on the server’s personal firewall and protecting UNIX-based services with iptables. Since using this method protects important administrative services like SSH, SNMP, and etc, from denial of service and/or any possible attacks, and moreover it is flexible for developers to have their own implementation, it is widely embraced by the industry and academic communities and several open source implementations have been released. The structure of this paper is organized as follows. In section 2, we examine the SSTCP method and review its advantages and drawbacks. In section 3, we will have a survey on existing implementations and proposed methods and study their benefits and flaws. We then present in detail our new proposal based on ESSTCP in section 4, and its method analysis in section 5. Finally, in section 6 we summarize and conclude the work.2. Spread-Spectrum TCP (SSTCP) 2.1. MethodologyIn this method, the client calculates the Authentic Port-Knock Sequence (APKS) and sends N TCP SYN packets to calculated port numbers. In the other side, there is a Silent Authentication Service (SAS) module on the firewall that listens for incoming knock sequence and if NM ≤ of them is received in correct order, it authorizes the client to send packets through the firewall on the desired port. In SSTCP, an infinite array of ports is generated by a one-way function based on a key (K) and time window value. Each new element of APKS in each time interval of “T” generated as []xFFFFtKSHAtA 0&)||(= (1) where “SHA” denoted as the hash function of SHA-1, || denotes concatenation, and “t” is the current time divided by T. They also propose a mechanism for synchronization on the first knock. In this mechanism, the received knock 1-4244-0910-1/07/$20.00 ©2007 IEEEis processed as shown in Figure 1. The first knock is compared with three generated knocks at time t-1, t, andt+1. if it was equal to one of them, the SSTCP considers an array of knock from A[start..start+N], and if the knocker is able to knock M ports out of N available ports, it is eligible to open its desired port. The detailed description of SSTCP is available in [1]. 2.2. SSTCP Analysis T2.2.1. Probability of guessing APKS.T Since this protocol is completely passive in the network and generates no traffic, the SAS module can be implemented on layer 2 (L2) devices such as hidden security boxes which have no IP address, example of which include layer 2 firewalls and IPS/IDSs. Furthermore, knocks have no specific signature to which may reveal the existence of this service on the firewall to the attacker. SSTCP is also resilient to the packet loss and the packet reordering. Its computation complexity is relatively low, and it is invincible to brute force attacks, since the probability of guessing is MMMNMNP16)1(1616SSTCPinAPKSGuessing211321123¸¸¹·¨¨©§−−⋅=¸¸¹·¨¨©§−−⋅=− (2) in which first term is for the first knock and second term is the probability of guessing other knocks. For example for the arbitrary values Tof N=10 and M=8 the probability would be-37102.1738 ×, which is small enough to be safe against brute force attacks. Barham et al. also addressed certain problems in SSTCP and discussed alternative approaches to mitigate these issues. One approach is using ISN (Initial Sequence Number) instead of the destination port, first to diminish the probability of guessing port knocks. Because the destination port field in TCP header is just 16 bits, while the sequence number field is 32 bits. Secondly, as some packets with specific destination ports might be dropped by the firewalls in between, ISN would be a good idea to be exploited. However, sending packets to a specific