How Chinese hacked Google, and why India should worry

The recent announcement by the United States giant search engine Google that it might withdraw from China made the headlines in world media. The Google decision highlighted the aggressiveness of the Chinese hackers who had been penetrating cyber fortresses like the Pentagon or the White House (as well as the PMO or the MEA in India!).

Claude Arpi spoke to Shishir Nagaraja, the co-author (with Ross Anderson) of The Snooping Dragon: Social malware Surveillance of the Tibetan Movement,  published by University of Cambridge Computer Laboratory in March 2009.

Shishir Nagaraja, currently associated with the Information Trust Institute of the University of Illinois (US), tells rediff.com, not only about the Google episode, but also his experience with the Office of the Dalai Lama in Dharamsala and the world of hackers, in general.

He believes that we have only seen the beginnings of the cyberwar, the 'war of tomorrow'. In the not-too-distant future, it will affect each one of us.

What according to you has happened with Google in China?

From what I could gather, they targetted some people connected to the Tibetan movement and some mainland activists.

The second aspect is that the infrastructure used by Google to carry out censorship in China was a part of the attack. Not very much has been made public by Google in this regard, so we can't be very sure.

Third, Google itself was a victim and they claim to have lost intellectual property. What we know for sure is that the email accounts of the Tibetan activists were read regularly from IP addresses in China.

. . .

What is new in these attacks? One reads that they were highly sophisticated?

No, it is the same old story. Nothing is new. It is the same thing that we wrote about [The Snooping Dragon report] or Greg Walton wrote about [Tracking GhostNet report]. Same thing!

The only new thing is that they have targetted Gmail addresses, but this was known to us. In fact, I had approached Google in September [2008] after the Office of the Dalai Lama's Representative in New York had got in touch with me; they had found out that somebody had maliciously configured their SMTP [outgoing mail] server so that it would forward all their emails to a certain Google account.

It is interesting because a lot of space is needed for this and Google has that space. Isn't it better to use something already available?

The Dalai Lama's Office [in New York] found out that even that [Google account] space had overflowed; they had not removed the wiretrap and the forwarded mail started bouncing from Google. It is then that they realised what was going on.

When I was approached, I advised them to talk to Google. Later, on their behalf, I informally talked to the person in Google responsible for investigating malicious activity. He said, 'You can put a formal complaint if you want, but there is not much that we can do.' This is the response that I got.

Some 30 other companies are said to have been attacked at the same time.

Yes, we had projected [such attacks] in our report. In fact, the theft of Google's IP is exactly the sort of attack we warned against. We had said that more and more people will use tactics pioneered by the 'Chinese hackers'. The attack this time is not different; the attack vector is the same, 'abuse of social trust'.

The [attackers] make your emails look like from someone you trust, not from a stranger. This is done by replaying past messages with minor modifications, and I expect the attackers will mature to the point of using victim input in real time to construct attack emails: for instance, by embedding malware into an attachment even as the victim composes a message.

Now for that part about why Google is behaving like this [threatening to withdraw]? There are no new technical reasons for doing so. There might be business reasons though. It is a tough market. They don't have a large share in China compared to their competitors.

This could be a face saving excuse or a bargain striking maneuver, I don't know.

. . .

Ultimately, to threaten to withdraw is good for their image?

It is favourable to their image. There is a lot of anti-China sentiment in the West. [Google's decision] plays into this, while giving them a good reason to withdraw, though I am not sure that they really want to withdraw, because political censorship climate has remained unchanged in China.

Ten/fifteen years ago, when they came to China, the Chinese government told them the same thing: you have to censor the Web. Today, Google says: 'We are negotiating with the Chinese government. We don't want to censor the Web!' The reasons stated now to leave the market were valid even when they entered the market.

Playing in a capitalist world, Google knew the rules of the game, and they were willing to play by it as long as they turned a profit. It was the same then, it has not changed, formally or informally.

Since 1989, the Chinese government is clear about their policy of censorship.

Could you tell us your experience with the Office of the His Holiness the Dalai Lama (OHHDL). Tell us about Snooping Dragon? It seems to have been interesting in the sense that you found an organisation willing to be openly studied, which not the case for governments, banks, Army, etc.

Yes, it is not usual, though since 2004 there have been some cases documented through Congressional hearings. In contrast, by agreeing to make the findings public, the Buddhists have shown themselves to be truly enlightened.

Though, from a political perspective, agreeing to make the subject public made a lot of sense [for them]. In the diplomatic battle between China and Tibet, the latter has always sought to portray an image of a victim set against an aggressive Chinese position.

It played [in favour] of their PR image. However, banks, governments and companies seek an image of 'nothing is wrong with our security'. But this is a rational explanation. I don't think His Holiness invited us with this in mind.

When we were invited to have a look, the OHHDL was not aware of the extent of damage being caused by the attacks much less being in a position to perform accurate diplomatic calculations.

. . .

It was quite bad?

Oh, yes, it was bad. Their electronic infrastructure was completely compromised. The bad news is that this attack can also be carried out on any usable computing infrastructure with very few exceptions, very few people believed in this assertion when our report came out, but the successful attacks on Google vindicate our position.

Could, for example, the attackers have known the position of the Dalai Lama's team before they went to Beijing for talks?

Very much possible if their position [for talks] was prepared and recorded on the computers. These days, the OHHDL is fairly tech savvy and use email and electronic storage for almost all their activities.

The Chinese stole detailed meeting notes, plans for school construction, basically any data sitting on an OHHDL computer was lifted. One of the most important was the refugee database.

It means all the registration details of all the Tibetans refugees who had fled to India.

The sys-admins took it offline as soon he realised that the attack was going on. Regarding the sys-admins, I have a lot of respect for the decisions they took. They took the right decisions and the level of response with speed and accuracy would be in line with the best trained sys-admins.

It is quite commendable really. They found a problem, and they asked experts for help immediately without trying to hide the problem or hoping it would go away. . . they wanted to find out. They found the best experts to help them. Usually the IT security culture of most organisations is to hide mistakes.

The sort of openness that the OHHDL has in matters of general policy as well in the management of their computer security is very commendable.

It is because of this culture [of openness] that they were able to discover the extent of surveillance going on. And for these reasons, we are much more aware of Chinese info-warfare capability.

To what extent the security holes have been closed, I am not sure. I don't think they have been closed. They are very much there and the attacks might be repeatable; it is a tough problem to solve.

. . .

If embassies or government offices can be attacked, one can presume that it is easier to penetrate relatively smaller office like the Dalai Lama's?

Yes, you are right. Similarly, if Google can be attacked, then most companies can be successfully targetted as well.

A news item mentioned that Tibetans would have stolen data from the Chinese, particularly the laptop of a lady-member of the United Front Works Department, the Chinese ministry dealing with the Dalai Lama's Envoys. Are you aware of this?

I don't know. I have not heard about this.

There is always the question of what constitutes proof in a computer security investigation. In the case of the OHHDL, the evidence I have used during the investigation, wasn't the IP address of the control server or similar information.

The main evidence comes from the fact that the Chinese foreign ministry used some of the intelligence information gathered from electronic surveillance and used it to apply diplomatic pressure on those invited to meet with the Dalai Lama.

When the Chinese foreign ministry showed full knowledge of OHHDL emails -- this constitutes strong evidence in my eyes -- it showed that there was Chinese government involvement at some level, although they might not have carried out the attack themselves.

The ownership of the attack is squarely with the Chinese government even if they might have 'outsourced' the attack to Chinese cyber-guerrillas.

In our report, we provided additional explanation on why we chose to point fingers at the Chinese government. We also considered other theories: who else could have been motivated to carry out this attack and why and if they had done it, what would be the evidence.

We have seen strong evidence of Chinese government involvement, and none to the contrary.

. . .

The media has recently dealt at great length on the so-called independent hackers and the role of the Chinese State.

In my mind, it is a little bit like guerilla warfare; a much sought-after alternative to conventional forces. Guerilla warfare provides plausible deniability to the sponsoring State. If you consider US-Iraq, US-Afghanistan, Pakistan-India or Israel-Palestine conflicts, we often see a model of 'guerilla warfare' playing out. It appears that such a model of warfare is gaining popularity.

If the quality of the fighters is very good on the 'open market', why not hire them instead of training your own and risking bad press.

Don't you think that China has this type of mindset to use these tactics while it is not present in India?

Well, there are documented cases of India's intelligence agencies using the underworld (Dawood versus Chhota Rajan, for example). But these are home affairs and have little to do with other countries.

In comparison, the Chinese use of guerrilla hacker networks is quite popular. Timothy Thomas has documented this quite well [it is referenced in the Snooping Dragon report].

The Chinese attacks on the OHHDL appear to have been carried out by semi-skilled amateurs. From the quality of the work, I can say that it was not a very skilled person, not a real expert. If they had experts on hand, then the situation would have really been different in terms of difficulty analysis.

This points to two things: one: analysis will get tougher in future as attacks get more sophisticated, and, second, if amateurs can carry out successful attacks on Google and OHHDL, then that signals a very real danger.

. . .

About Chinese 'experts': do you believe that many of them have been trained in the US or the West and later returned to China?

Possibly! But there is no need for a good hacker to be trained in the US. People with good computer skills are very much there in countries like India, Pakistan or China. Some very, very skilled people might not even have had elementary education.

The Chinese recently closed a 'hacking' school in Hebei province. Is it eyewash, or will it make a difference?

[These days] there are loads of resources online, so closing one school won't make a difference for the same reason that closing a terror school hasn't made a difference.

If someone wants to learn, it does not take much effort. It is important to understand that the main innovation is not technological, it is a psychological one. The entire computer industry has progressed technologically, but computer security is not a technology issue.

Technologies are fine, they are there. The question is the human link. The way humans interact with computer security is poorly understood by software engineers.

The current technology does not consider humans as they are: humans are fitted into a user model of how they are 'supposed' to be. Each time there is a security problem, security experts are quick to point to the user's fault! The user did not do this or that! This mindset has to change.

Technology needs to understand and accept user behaviour and provide security assurances with this in mind. We should accept people as they are, accept the diversity in human behaviour, there is no point in writing manuals and designing secure systems for somebody else.

The users are not going to do change, so user education is the wrong place to spend security budget.

. . .

In their White Paper of Defence, the Chinese strategy has undergone a shift from 'active defense', (never attacking someone first, but being ready to respond if attacked) to 'active offense'. Don't you think that a nation practicing this will always be a step ahead of its opponent?

As usual, computer security is quite asymmetric. It takes less to attack than to defend. You have only to find one hole to be successful in attack, while defence has to plug all the holes.

For this reason, it appears that attacking is easier than defending, computer systems or physical world security.

Recently, an article in the Indian Press affirmed that the National Technical Research Organisation which deals with cyber attacks in the government pretends that their Rapid Action Group can tackle an attack in less than 90 minutes. What are your views on this?

Assuming they mean 'any' intrusion, it is highly, highly unlikely to be true. If it was true, it would be a five-star research contribution, probably worth a Nobel Prize.

Instead, if they are claiming that the exact same attack would be detectable that's straight forward but close to useless in defending against future attacks (they won't be the same as past attacks).

Attacks don't repeat the same way. . . why should they? They always evolve. To prove that nobody can steal an organisation's data, you have to prove that every hole has been closed.

[However] there are not just bugs in software; there are also bugs in human operation. For example the attack on the OHHDL was not due to a computer bug, the software defects were there, but they were incidental to the attacks.

When humans authenticate emails, they do so based on socio-cognitive signals based on the text of the email. It is a highly sophisticated pattern analysis-based authentication mechanism that is used by humans.

The attackers found a way to beat it by simply replaying the text. In this type of an attack, detectability is very low. If the attacker decides to intrude and stay around your network, it might take a couple of years before he/she is detected, [he can remain dormant].

In the case of the OHHDL, they were probably there for a year or so. The attackers were detected, because they increased the frequency of attacks way too much. They made two mistakes: one they replayed emails too many times, and second, they showed that they knew some information that they could have not known without spying.

But the attackers will learn and the second generation of social malware attacks will be more covert. Will we detect them? Unlikely! In half an hour? Very, very unlikely!

When the Pentagon or the White House have been penetrated [in the past], it took [sometimes] years to find out. They are ways to remain covert, attack covertly (no replays), transmit covertly (using covert channels/'96 there are lots of them).

Presence of attacks on OHHDL could be found out [relatively easily]. But if they deployed covert communication over the Internet to transfer stolen information, then they can remain virtually undetectable for a very long time.

. . .

Recently, DefExpo India 2010 was held in Delhi. The Indian government is planning to spend Rs 50,000 crore (Rs 500 billion) in military hardware, don't you think that it is not the 'war of yesterday'?

Oh, yes! Absolutely! What you mentioned is conventional warfare. Now we are speaking of guerilla warfare. A significant national security risk to India lies in the area of computer security which can't be addressed with Sukhois.

With the increasing reliance on computer networks, India's information infrastructure is growing rapidly. The budget for computer security has to increase too.

There is a very real risk that China has control over significant parts of the government's computer infrastructure. Military capability will mean little if the enemy has high quality intelligence.

Supremacy in information security is crucial, for economic security reasons too. For example, how to protect IP from India's software industry from being stolen? Social malware can be used to steal software.

Another example involves injecting false data into accounting systems. Each company has an accounting system which is automated using computers. Social malware can be used to infect a majority of the computers of an accounting system.

With banks having a hard time coping with 1 per cent of customer machines being infected, how can a company run an accounting system with 50 per cent of its machines being compromised?

The scale of such economic fraud could run into hundreds of millions of dollars. And it is increasing, even as we speak.

We all need security against social malware attacks. Political organisations could be hit and have their political secrets revealed. Consumers and business organisations will be hit by accounting frauds.

In today's economic climate, such frauds might be enough to put small companies out of business. Today, even for a small company, you can't do your accounts manually. . . if a malware introduced false transaction amounts of Rs 10,000 or Rs 15,000, this won't even be noticed until it is too late and money has been siphoned off using Western Union.

If the behaviour of banks in the case of ATM frauds is anything to go by, then banks will simply dump the liability on the end users saying 'it is your fault; the malware was in your computer.'

The negative fallout will always have to be taken by the customers who do not have the means to defend themselves. I foresee that we will witness new instances of social malware attacks, targetting businesses and individuals in the near future.

. . .

Tell us something about your project in India

I will move to India shortly. I will take a position of Assistant Professor at the IIIT Delhi and, with a group of three colleagues, will start a Security Group conducting research and teaching in computer security.

We have a Master's and a PhD programme. My first priority will be to carry out a comprehensive analysis of the scale of computer crime in India. Today, this research is carried out by people from outside [India].

To carry out defensive actions, we have to know the scale of exposure to [computer piracy]. What we did for the OHHDL, we will do for various companies and governmental organisations. It means high level audits. It is a lot of work. All the information is scattered today, it may take a while to get the data, analyse it, publish the results and take remedial measures.

The government can't do everything, but it can start programmes to improve computer security for the public.

Thank you.