University of St Andrews



Kruptos Research Project by David Llamas


Kruptos Research Project: Covert Channels and Steganography in Computer Networks

Kruptos means "hidden", "secret" in ancient greek, which are concepts closely related to this research.

A covert channel is a communication channel that allows two cooperating processes to transfer information in a manner that violates the system's security policy. It is thus a way of communicating which is not part of the original design of the system. The use of the covert channels and steganography in public computer networks can prove an effective means of information hiding and secret communication. With the widespread adoption of the Internet the TCP/IP suite of protocols have become pervasive, and therefore an attractive target for the use of these techniques. Opportunities for the creation of covert channels exist at all layers in the TCP/IP protocol stack.
Information hiding techniques can be used by criminals and terrorists to communicate over covert channels within digital networks and can be used to overcome firewalls and most other forms of network intrusion detection and prevention systems. In fact, most systems can detect hidden data in the payload, but struggle to cope with data hidden, for instance, in the IP and TCP packet headers, as no discernable patterns from which hidden data can be analyzed and detected are produced. This work is clearly relevant to governments, defence departments and intelligence agencies, as well as to all organisations that can be affected by academic or industrial espionage.

University of St Andrews ID Card
David Llamas

Kruptos is sponsored by the Defence Science and Technology Laboratory -DSTL- a part of the Ministry of Defence -MoD- of the UK.
Network Traffic Oscilloscope & Covert Channels in Computer Network Protocols Detector

Kruptos Research Project

A Potential Covert Channel Manipulation Detector based on the IPv4 header ID field analysis has been developed in C# on Microsoft .NET Framework using Visual Studio IDE, which includes the following features:
  • Automatic traffic request engine, from which networks packets are required to be sent from a specified host by requesting the default web page once and again within a loop.
  • Incoming traffic monitor, which permits to specify what fields from the IPv4 header are required to be shown.
  • A connection performance meter, to know how many packets per minute are received from that connection.
  • A detection algorithm control panel. Potential covert channels detection is based on the position within a range of values in an internal counter called Covert Channel Meter, specified by a Minimum and Maximum value. Minimum value means low risk of potential covert channel manipulation, and Maximum value means high risk. Each time a network packet is received the value of the IPv4ID field is compared with the previous one. If the difference is within the Jump Range value, what is specified in the Increment Count field is added to the Covert Channel Meter. Otherwise, what is specified in the Decrement Count is subtracted from the Covert Channel Meter. Due to sometimes it has been observed breaks within the sequence where the value in the IPv4ID field is set to a lower value (or 0), an option to take into account or ignore Negative Jumps have been implemented. By adjusting these parameters, potential covert channels based on the manipulation of the IPv4ID field can be detected. This algorithm is based on the current implementations of the TCP/IP drivers in the major important operating system in the market.
  • Data logger, that records the information received from the two connections that can be managed simultaneously from the application.

Covert Channels Detector

In order to conduct this work, a research about how the IPv4 header ID field is implemented in the major operating systems in the market has been carried out. These are the results from the network traffic oscilloscope developed for Kruptos:
  • The IPv4ID generation algorithm is initialized in Linux associated with the TCP socket, and it's an incrementing sequence from there.
  • In SUN Solaris, the IPv4ID is initialized with a secure random number. IDs are simply incrementing from there or randomized each time is needed. It uses distinct counters for each destination IP address.
  • The IPv4ID algorithm in OpenBSD uses a linear congruential generator, rekeyed every 3 minutes (or after 30,000 IDs have been generated, whichever is sooner).
  • In MS Windows, the IPv4ID increases monotonically until reaching the maximum value of 65533, and then starts again from 0.

Linux 2.4.x - http://www.st-andrews.ac.uk

SUN Solaris - http://www.ebay.co.uk

OpenBSD - http://www.openbsd.org

MS Windows - http://www.microsoft.com

Webwall - http://www.dstl.gov.uk

Reverse Proxy Server - http://kruptos.inchcolm.org

Random? - http://www.mossad.gov.il

Unknown

What it can be seen from this is that the major implementations in the market follow the legal framework of the protocol, which is that the sender must choose the identifier to be unique for this source for the time the datagram (or any fragment of it) could be alive in the internet, but in a different way that is suggested by RFC 791, which is that the sending protocol module should keep a table of unique Identifiers to manage the exclusivity of that identifier. Instead of that, current implementations sequence the number.
Distributed-System Monitor for Covert Channels Analysis at Planetary-Scale

Planetlab

It consists of software components that work separately and concurrently. A distributed-system monitor is divided into a number of layers. Each layer makes use of the services provided by the lower layers and extends the available facilities to the upper layer. The layers are briefly introduced:
  • Observation: This layer gathers raw data on individual components of the system. Generally, each component may have an observer designed specifically for it. Thus there may be several observers located on different subsystems.
  • Collection: This layer collects data from various observers. It is possible to have more than one collector on large systems. Collectors may coordinate the activities of observers across the network and sometimes be responsible for store and forward activities for collectors situated on other nodes.
  • Analysis: This layer analyzes the data gathered at various collectors. It may consist of various statistical routines to summarize the data characteristics. Simple analysis such as counting of events is done most efficiently in the observer but more complex statistical calculations are performed here.
  • Presentation: This component of the monitor deals with human user interface. It produces reports, displays, and alarms. The presentation layer may be integrated within another tool that may manage some part of the system.
  • Interpretation: This component refers to an intelligent entity (human or expert system) that can make meaningful interpretations of the data. This generally requires multiple rules and trend analyses. Simple threshold-based alarms may be considered part of the presenter rather than of the interpreter which usually requires the application of more sophisticated rules.
  • Console/Control: This component provides an interface to control the system parameters and states. Strictly speaking console/control is not a part of the monitor. However, the monitoring and control functions are often used together and it is desirable to allow system control as well as system observations facilities to be used together.
  • Management: The entity that makes the decision to set or change system parameters or configurations based on interpretation of monitored performance is called the manager. A software manager component exists only in monitors with automated monitoring and control facilities.

Distributed-System Monitor for Covert Channels Analysis at Planetary-Scale

The distributed-system monitor consists of multiple components from each of the layers listed above. There is a many-to-many relationship between successive layers. A single collector may gather data from multiple observers. Many analysis layers may use data gathered by a single collector. Several collectors may be acting with a single observer, etc. Interpretation, Console/Control and Management layers can also be automated but they are often done by humans. For this reason the distributed-system monitor here is designed taking into account layers from the Observation to the Presentation, which includes a console for managing the whole system. The components of the distributed-system monitor are able to operate at any location on the Internet:

  • Observers analyse traffic from specific segments of the network.
  • Collectors are distributed taking into account, for instance, traffic payload capabilities.
  • Analysers work in a specific mode according to a particular strategy within a scenario.
  • Presenters show the system management and results in the appropriate view based on the context.

All components of the distributed-system monitor have been developed in Java and Python using NetBeans IDE and deployed at PlanetLab, an open platform for developing, deploying and accessing planetary-scale services.
School of Computer Science

School of Computer Science

The School of Computer Science at the University of St Andrews (rated 5A, outstanding international research) is a centre of excellence for computer science teaching and research with staff and students from Scotland and all parts of the world.

David Llamas at the School of Computer Science, North Haugh, University of St Andrews

It has international reputation in the areas of fundamentals of computer science, networks and distributed systems, artificial intelligence and software systems engineering.

The School of Computer Science at the University of St Andrews is 2nd in UK and 1st in Scotland

University League Table Computer Science

Reference: The Guardian

The School is a member of the Scottish Informatics and Computer Science Alliance (SICSA) whose aim is to promote and improve computer science research in Scotland.
University of St Andrews

University of St Andrews

Founded between 1410 and 1413, the University of St Andrews is the oldest university in Scotland and the third oldest in the UK. It is situated in the small town of St Andrews, in Fife (Scotland, UK) and is one of the most prestigious universities in the UK. It is often seen as an alternative to England's universities of Oxford and Cambridge.

The University of St Andrews is 3rd in UK and has the highest completion rate in Scotland

University League Table

Reference: The Guardian

The University of St Andrews is Scotland's top rated research institution and one of the leading research-intensive universities in the world (94% of St Andrews' research activity is internationally recognised and 60% is world leading or internationally excellent).
The Town of St Andrews

St Andrews (Scottish Gaelic: Cill Rìmhinn) is a small town of around 18,000 people situated on its own bay beside the North Sea on the east coast of Scotland about 50 miles north of Edinburgh, Scotland's capital city. The medieval layout of the town centre remains intact: beautiful old stone buildings, quaint narrow lanes and broad, tree-lined streets contribute to the unique, safe and tranquil atmosphere of St Andrews.
St Andrews has acquired the name "home of golf" for two reasons. First, the Royal and Ancient Golf Club, founded in 1754, exercises legislative authority over the game worldwide except in the U.S. and Mexico. Second, the beautiful links (acquired by the town in 1894) is the most frequent venue for The Open Championship, the oldest of golf's four major championships.

Map of St Andrews

About the middle of the tenth century, Saint Andrew became the patron saint of Scotland, and his feast day is on November 30th. The Saltire (or "St. Andrew's Cross") is the national flag of Scotland.