One of the cornerstones of policing is the ability to identify perpetrators to bring them to justice. For cybercrime, this is a difficult task, due to the multitude of ways to create anonymity over the Internet.
As a trivial example, imagine an attacker sending messages to a victim through a third computer called a proxy. After the attack, the proxy is wiped of all data, losing the ability to directly attribute previous messages.
Botnets are a more complex example of this, where a large number of computers conspire to remove the ability to trace messages to the original attacker. Tor is another example, with the same technology simultaneously used both to support freedom against oppression and to enable cybercrime to happen anonymously.
With the number and quality of anonymisation technologies available, it would seem that attack attribution is a lost cause. Direct attribution, the tracing of a message across networks to its origin, is exactly the problem that is "solved" by these technologies.
In many cases, the only way attribution through networks such as Tor has been through loopholes or security flaws, which are quickly overcome once discovered.
There exists another possibility for attribution, known as indirect attribution. In indirect attribution, the methods and attributions of an attack help build a model which is then used to link attacks to a suspect.
One example of this is the use of language and grammar in phishing attacks - such language can be used to profile an attacker, obtaining information such as age, gender and first language. This type of attribution is much more difficult to overcome, as it forces attackers to remove features from their attack which may be necessary to the attacks success.
In phishing attacks, this may require the phisher to remove forceful language such as the infamous "You must activate your account within the next 24 hours". Such information, the exact way in which the message is conveyed, improves the effectiveness of indirect attribution methods.
Language is not the only way indirect attribution can occur; an attacker trying to perform a brute force attack on a password may use a unique rainbow table which may later be discovered in another attack.
Indirect attribution at the cutting edges of cybercrime research, with research into enabling methods being performed around the world.
Indirect attribution will pose a major threat for cybercriminals in the next decade. It creates a fine line for attackers; do they increase their customisation of the attacks and risk being personally identified or do they use standard attacks and risk being blocked by simpler pattern recognising algorithms?
Indirect attribution also poses a challenge for forensic investigators and administrators. The information used to attribute an attack may not be immediately obvious, forcing a larger requirement to store more data about an attack.
When did it happen? Which specific attack messages were sent? Which router did those messages come from? What was the character encoding of the email? All these questions, and many more, may give the answer to attack attribution in the future.
I welcome your thoughts, input and feedback via comment.