Security Policy, Terror and Freedom: An Analysis of Threats and of Society's Reactions

The Era of Overreactions

The attacks of 11 September 2001 had a profound influence on the world of information security and on global politics. As Osama bin Laden remarked, "Al-Qaeda" spent about 500 thousand dollars on organizing the attacks, whereas the USA, as a result of the events themselves and their consequences, lost, by the most modest estimates, more than 500 billion dollars. Modern terrorism largely works by provoking the state into excessive and irrational reactions.
This phenomenon gives rise to a number of critical questions for society and for security engineers:

1. The political question. Are Western open societies, with democracy and a free press, vulnerable precisely because of their openness? Responses to terrorist threats often lead to authoritarian measures: fingerprinting at airports, the introduction of identity cards, total surveillance and detention without trial, which undermines centuries-old traditions of the rule of law.
2. The economic question. Colossal sums are spent on security measures with zero or minimal effectiveness. For example, the USA spent more than 14 billion dollars on the screening of passengers at airports without catching a single terrorist. At the same time, reinforcing the doors of pilots' cabins, which could have eliminated the main threat, would have cost a fraction of that sum; however, "invisible" measures bring no political dividends.
3. The problem of the "security military-industrial complex". After 11 September there arose a rapid growth of corporations lobbying their interests in the security sphere, which resembles the defense industry of the Cold War era.
4. The manipulation of technical arguments. Technical arguments are often used to justify bad policy. For example, in the UK the period of detention of suspects without trial was extended, argued by the necessity of decrypting data from computers. In reality the problem lay in a banal shortage of interpreters from Somali, which could have been solved differently.
5. The domino effect. Policy in the field of cyberspace has shifted toward surveillance, censorship and export control, which directly affects civil liberties and privacy.

Terrorism: The Evolution and Nature of Political Violence

Historical Context

Political violence is not an innovation; anthropologists find evidence of tribal warfare among early humans and chimpanzees. Terror was used to suppress subjects by the Maya, the Incas and William the Conqueror. In later history one may recall Guy Fawkes (the Gunpowder Plot of 1605) or the Irish Republican Army (IRA).
It is remarkable that the boundary between a terrorist and a statesman is often blurred. Many leaders who used violence in the struggle for their goals subsequently became heads of state: Gerry Adams, Menachem Begin, Charles de Gaulle, Jomo Kenyatta and Nelson Mandela.

The Causes of Political Violence

The first piece of good news is that the general trend in terrorist violence is steadily declining. A study by Paul Collier and Anke Hoeffler carried out for the World Bank analyzed wars from 1960 to 1999 in order to determine what drives conflicts: "grievance" (inequality, the absence of political rights, ethnic divisions) or "greed" (the economic viability of the rebels).
The data convincingly show that grievances play an insignificant role. Insurgencies are determined by whether they can be financed. As Cicero remarked long ago: "Endless money is the sinews of war".

• The IRA campaign continued thanks to the support of the Soviet bloc and of Irish emigrants in the USA. When this stream dried up, the weapons were laid down.
• The conflict in Sierra Leone was fueled by "blood diamonds".
• The insurgency of the Tamil Tigers in Sri Lanka was financed by the diaspora in the USA and India.
• "Al-Qaeda" was sponsored by wealthy donors from the countries of the Persian Gulf.

The conclusion for security: the most effective way of combating insurgents is to cut off their financial flows.

The Psychology of Political Violence

People are inclined to ignore probabilities when emotions come into play (the affect heuristic). Horrifying events such as terrorist attacks make society overestimate risks.
An experiment carried out with 22 municipal judges in Tucson (Arizona) showed the influence of "mortality salience". The judges were asked to set the amount of bail for a prostitute suffering from drug addiction. Half of the judges were beforehand asked questions forcing them to think about their own death. The result: the judges who had been reminded of mortality set bail averaging 455 dollars, whereas the control group set only 50 dollars. Awareness of the finiteness of life makes people adhere more strictly to cultural norms and defend their worldview more aggressively. The attacks of 11 September, by reminding people of death, provoked a surge of religiosity, patriotism and, as the reverse side, prejudice.

The Role of Political Institutions

Why do governments take irrational decisions? The answer is given by "public-choice economics", founded by James Buchanan (a Nobel laureate). Politicians strive to maximize their chances of re-election, and officials to expand their empires and budgets.
For example, Britain's initial reaction to 11 September was to increase the budget of MI5. However, this brought no "political meat" to the security industrial complex. As a result, the funds were redirected to the grandiose and expensive project of introducing national identity cards. Similarly, the invasion of Iraq was regarded by many analysts as a way of providing work for the Pentagon and the defense industry, which had been left with nothing to do after the successful but short war in Afghanistan.

The Role of the Press

The mass media live by the principle "If it bleeds, it leads". Editors want to sell more newspapers and to broadcast the most frightening versions of events, thereby intensifying panic and allowing politicians to dictate the agenda, keeping their opponents on the defensive.

The Democratic Response and the Education of Society

Despite the gloomy trends, democratic societies are capable of learning from their mistakes. The public reaction to the explosions in the London underground on 7 July 2005 was far more restrained than the reaction to 11 September in the USA — largely thanks to the experience of a quarter-century of struggle against the IRA.
The historian John Mueller shows how the attitude of American administrations to terrorism changed:

• Jimmy Carter inflated the hostage crisis in Iran, which gave him an initial rise in his ratings but in the end cost him the presidency.
• Ronald Reagan mostly ignored provocations, but succumbed to temptation in the case of the Lebanese hostages (the "Iran-Contra" scandal), after which he quickly restored his rating by distancing himself from the mistake.
• George Bush used a policy based on fear, which in the end led to a fall in his popularity.

Voters over time see through the fear. Politicians who try to use terrorism to intimidate and to obtain votes face public rejection in the long run.

Surveillance

One of the side effects of the events of 11 September 2001 was the colossal increase in the scale of technical surveillance, both by wiretapping and through the data mining of commercial data by government agencies. Recent disclosures of the facts of unlawful surveillance in a number of countries, along with the differences between the approaches of the USA and Europe to confidentiality, have politicized these questions. Wiretapping was already an acute topic in the 1990s, and millions of words have been written about it.

The History of Government Wiretapping

Rulers have always tried to control communications. In classical times couriers were checked at customs, and from the Middle Ages many kings either ran the postal monopoly themselves or granted it to favorites. The "Black Chambers" — the services for the interception and decryption of mail in the early modern period — are described in David Kahn's fundamental history.
The invention of electronic communications provoked a defensive reaction. In most of Europe the telegraph service was created as part of the Post Office and was always state-owned. Even where this was not so, regulation was so rigid that the growth of the industry was severely braked, which left America with a clear competitive advantage. The multitude of national rules, which sometimes conflicted with one another, irritated everyone so much that in 1865 the International Telegraph Union (ITU) was created. However, this did not satisfy everyone: in Britain the telegraph industry was nationalized by Gladstone in 1869.
The invention of the telephone further intensified the government's interest in surveillance and the resistance to it — both legal and technical. In 1928 the US Supreme Court, in the case of Olmstead vs United States, ruled that wiretapping did not violate the provisions of the Fourth Amendment on unlawful search and seizure, since there had been no physical intrusion into the dwelling. Judge Brandeis famously entered a dissenting opinion. In 1967 the court changed its position in the case of Katz vs United States, ruling that the amendment protects people, not places. In the following year Congress legalized federal wiretapping (in "Title III" of the Omnibus Crime Control and Safe Streets Act) after testimony about the scale of organized crime. In 1978, after the investigation of the abuses of the Nixon administration, Congress passed the Foreign Intelligence Surveillance Act (FISA), which regulates wiretapping for the purposes of national security. In 1986 the Electronic Communications Privacy Act (ECPA) weakened the provisions on Title III warrants.
By the beginning of the 1990s the spread of deregulated services — from mobile telephones to call forwarding — had begun to undermine the ability of the authorities to carry out wiretapping, to which technical innovations also contributed, such as out-of-band signaling and adaptive echo suppression in modems. Therefore in 1994 the Communications Assistance for Law Enforcement Act (CALEA) was passed, which obliged all telecommunications companies to make their networks accessible for wiretapping in ways approved by the FBI. By 1999, there had been 2,450,000 line-days legally wiretapped in the USA, which on average amounted to just over 1350 simultaneous wiretaps authorized by judicial warrants.

DES as an Instrument of Policy

The NSA promoted the DES standard (with its 56-bit key) not simply as an industry standard, but as a deliberate political instrument. The idea was to hinder foreign governments from introducing strong cryptography. The NSA created the illusion that it possessed the computing power to break DES, which forced other countries to continue using vulnerable rotor machines, whose traffic the NSA could freely intercept and read.

The Growing Disputes Around Traffic Analysis

However, the main mass of police communications intelligence in developed countries comes not from the interception of content, but from the analysis of telephone bills and other data about connections. Criminals go to great lengths to hide their signals in innocuous traffic, using such methods as disposable mobile telephones and the hacking of PBXs, while the police, for their part, develop methods of tracking networks of contacts.
This is not new. Rulers have long used their control over the postal services to track the correspondents of suspects, even when the letters themselves were not opened. The invention of the postage stamp in 1840 was a step forward for confidentiality, since it made sending a letter anonymously much easier. Some countries so feared the threat of sedition and defamation that they passed laws requiring a return address to be indicated on the back of the envelope. The development of the telegraph, by contrast, was a step forward for surveillance: since messages were registered by sender, recipient and number of words, it was possible to compile traffic summaries which, as turned out, were an effective indicator of economic activity. The First World War showed the combatants the value of the intelligence that could be obtained by monitoring the volume of enemy radio traffic, even if it could not be decrypted. Subsequent conflicts only consolidated this.
Traffic analysis continues to provide the main mass of police communications intelligence. For example, in the USA in 1998 (the last year for which comparable statistical data were available) 1350 applications for wiretaps were approved, while 4886 judicial subpoenas were issued (plus 4621 extensions) for the use of pen registers (devices recording all the numbers dialed from a particular telephone line) and 2437 subpoenas (plus 2770 extensions) for trap and trace devices (which record the calling line identifier of incoming calls, even if the calling subscriber tries to block it). In other words, there were more than ten times as many subpoenas for data about connections as for the interception of content. Moreover, these data were even more incomplete than the wiretap data. The trend of recent years — even before 11 September — has been to move over to using subpoenas to obtain itemized billing from the databases of telephone companies, rather than obtaining pen register data directly from the exchange (to which CALEA contributed). For example, Bell Atlantic reported that over the period from 1989 to 1992 it had answered 25,453 subpoenas or judicial warrants for the itemized billing of 213,821 of its customers, while NYNEX reported that it had processed 25,510 subpoenas concerning an unspecified number of customers in 1992 alone. On a national scale this suggests that perhaps half a million customers annually have their records seized, and that traffic data are collected in roughly a hundred times as many cases as wiretapping is applied.

Unlawful Surveillance

In 2006 it emerged that the rumors had been true. The company AT&T had indeed been handing over to the NSA records of the calls of millions of Americans; the agency's goal was to "create a database of every call ever made within the national borders", so that it could map the entire social network of the USA for the "War on Terror". Apparently these data were now being collected for 200 million customers of AT&T, Verizon and BellSouth, the three largest telephone companies in the country. The program began shortly after 11 September. The company Qwest did not cooperate, because its chief executive at the time, Joe Nacchio, did not believe the NSA's assertions that Qwest did not need a judicial warrant (or FISA approval). The NSA put pressure on Qwest, threatening to deprive it of classified contracts, and Qwest's lawyers asked the NSA to submit its proposal to the FISA court. The NSA refused, stating that the court might not agree with them.
It later emerged that the NSA had been pressing Qwest to hand over the data even before 11 September. In October 2007 the Democratic senators received further confirmation when Verizon admitted to them that it had handed over to the FBI data about the second-generation calls of its customers under national security letters 720 times since 2005; and in November 2007 the Washington Post revealed that the NSA had wiretapped a multitude of purely domestic telephone calls and traffic data, and had also intercepted traffic at AT&T's peering node in San Francisco in order to obtain access to internet traffic.
Records of telephone and computer services may be provided to bodies other than law enforcement ones in accordance with 18 USC 2703; thus, for example, it can be seen that Virginia and Maryland plan to use mobile telephone tracking data to monitor congestion on the Capital Beltway. The use of traffic data for marketing purposes was also expressly envisaged by Congress when adopting this law. However, the growing availability of mobile telephone records has now made them available to criminals too, which allows gangsters to track targets and to find out whether any of their accomplices has telephoned the police.

Access to Search Queries and to Location Data

One problem is that data about connections and content are becoming ever more interconnected, since what is content at one level of abstraction is often connection data at another. People may think that a URL is simply the address of a page to be fetched, but a URL such as http://www.google.com/search?q=marijuana+cultivation+UK contains the terms entered into the search engine as well as its name. Obviously, some policemen would like to obtain a list of everyone who sent such a query. It is equally obvious that providing this kind of data to the police on a large scale would have a chilling effect on online discourse.
In the USA the Department of Justice sent a judicial subpoena to a number of search engines demanding that they hand over two full months of search queries, as well as all the URLs in their index, stating that these data were needed to support its assertions that the Child Online Protection Act does not violate the constitution and that filtering can be effective against child pornography. AOL, Microsoft and Yahoo quietly complied, but Google resisted. The judge ruled that the Department would not receive the search queries, but only a random sample of the 50,000 URLs that it had originally requested.
In Britain the government tried to define URLs as traffic data when it was pushing the RIP bill through parliament, and the news that the police would have unlimited access to the URLs entered by every user (their clicks) provoked a public outcry against "Big Browser", and the definition of communications data was cut back. For general internet traffic this means IP addresses, but it also includes email addresses. All this may be demanded by a simple notice from a senior police officer.
There are even more subtle points connected with the telephone system. In Britain all information about the location of mobile telephones is considered traffic data, and officials obtain it easily; but in the USA the Court of Appeals ruled in 2000 that when the police obtain a warrant for the location of a mobile telephone, it is sufficient to indicate the cell in which it is active, and that the requirement of triangulating the device (the interpretation the police wanted) would violate privacy. In addition, even cell-level location information would not be available under the lower standards applied to pen register subpoenas. Subpoenas were also found insufficient for post-cut-through dialed digits, since there is no way of distinguishing in advance the digits dialed for routing calls from the digits dialed to access or provide information. In practice this means that if a suspect goes into a 7-Eleven store and buys a telephone card for a few dollars, the police cannot obtain a list of his calls without a full wiretap warrant. All they can obtain by subpoena is the digits he dials to reach the operator of the telephone card, and not the digits he dials later to be connected.

Data Mining

The analysis of call data is only one aspect of a far broader problem: the matching of law enforcement data, that is, the processing of data from a multitude of sources. The earliest serious example of the use of data from several sources known to the public occurred in Germany at the end of the 1970s, in the search for the safe houses used by the Baader-Meinhof terrorist group. Investigators looked for rented apartments with irregular peaks in utility consumption, paid for by remote bank transfer from a series of different places. This worked: a list of several hundred apartments was obtained, among which there were several safe houses. The tools for performing this kind of analysis are now supplied together with a number of products used for traffic analysis and for managing large police investigations. The extent to which they are used depends on the local regulatory climate; in Britain there were disputes over police access to the databases of prescriptions issued by pharmacists, while in the USA doctors are concerned about how often personal medical information is summoned by subpoenas from insurance company investigators. There are also practical constraints imposed by the cost of understanding the multitude of proprietary data formats used by commercial and government data processors. But it has become normal for the police to have access at least to utility data such as electricity bills, which are analyzed in the search for marijuana growers.

Surveillance Through Providers: Carnivore and Its Descendants

One of the major recent developments has been intrusive surveillance through internet service providers (ISPs). Intercepting internet traffic is more complex than it formerly was with voice communication; there are a multitude of obstacles, such as the temporary IP addresses assigned to most customers, and the increasingly distributed nature of the traffic. In the old days (say, in 2002) a provider might have modem racks and a local network where a wiretapping device could be placed; now many customers connect through DSL, and providers use switched networks that often have no obvious place to install an interceptor.
In many countries there are laws requiring providers to assist wiretapping, and the usual way for a large provider to do this is to have equipment already installed that splits the target network so that copies of the packets of interest are sent to a separate classified network with wiretapping equipment. Small providers often have no such capabilities. At the end of the 1990s the FBI developed a system called Carnivore, which they could bring to small providers when a wiretap was required; this was a PC with software that could be configured to record a suspect's email or his web surfing, or any other traffic specified in the warrant.
It became controversial in 2000, when a provider challenged it in court; it had been used to record email headers as traffic data, without a wiretap warrant. At present we have FOIA disclosures about the FBI's current wiretapping network, DCSNET, which is very effective — it allows agents remotely and almost instantaneously to obtain access to the traffic and content of participating telephone companies.

Signals Intelligence Against Foreign Targets and Its Effectiveness

The main mass of communications intelligence, whether wiretapping, traffic analysis or other methods, is conducted not for law enforcement purposes but for foreign intelligence. In the USA the main agency responsible for this is the National Security Agency (NSA), which has enormous facilities and tens of thousands of employees. While law enforcement agencies have 150–200 active wiretaps at any moment, the NSA completely eclipses these figures. The situation is similar in other countries; the British Government Communications Headquarters (GCHQ) has thousands of employees and a budget of about a billion dollars, while for many years a single policeman at New Scotland Yard was engaged in administering all the police wiretaps in London (and did other things too).
Information gradually leaked out about the scale and effectiveness of modern signals intelligence operations. David Kahn's fundamental history laid the foundation, describing much of what happened before the beginning of the Second World War. An anonymous former NSA analyst, later identified as Perry Fellwock, disclosed the scale of the NSA's operations in 1972. "NSA collection is complete", he wrote. "It covers what foreign governments do, what they plan to do, what they did in the past: which armies move where and against whom; which air forces move where and what their capabilities are. The NSA really has no limits. Its mission extends from calling in a B-52 in Vietnam to monitoring every aspect of the Soviet space program".
While Fellwock's motive was opposition to the war in Vietnam, the next major whistleblower became a British codebreaker of the Second World War, Frederick Winterbotham, who wanted to write memoirs about his wartime achievements and, since he was dying, was not worried about prosecution. In 1974 he disclosed the success of the Allies in breaking German and Japanese cipher systems during that war, which led to the appearance of a multitude of other books on Second World War signals intelligence. After this there was a slow stream of revelations from investigative journalists, quite few of whose sources were concerned about the abuse of facilities by officials tracking domestic political groups. For example, the whistleblower Peg Newsham disclosed that the NSA had unlawfully wiretapped a telephone call made by Senator Strom Thurmond.
James Bamford gathered quite a lot of information about the NSA from open sources and by talking to former employees, while the most substantial recent source of information about the organization and methods of US signals intelligence and that of its allies became the book assembled by the New Zealand journalist Nicky Hager after the New Zealand intelligence community disobeyed the order of its prime minister to downgrade the level of intelligence cooperation with the USA.
The picture that emerged from these sources described a worldwide system for the collection of signals intelligence, Echelon, jointly run by the USA, Britain, Canada, Australia and New Zealand. Data, faxes and telephone calls are collected at a large number of nodes, which range from international communication cables that land in the participating countries (or are covertly intercepted under water), to monitoring the traffic to and from commercial communications satellites and special intelligence satellites that intercept traffic over hostile countries, to wiretapping points in the embassies of the participating countries. The collected traffic is searched in real time by computers known as dictionaries, according to such criteria as the telephone numbers or IP addresses of the sender or recipient, as well as keyword searching in the content of email. These search criteria are entered by the intelligence analysts of the participating countries; the dictionaries then collect the traffic satisfying them and send it back to the analyst. Echelon apparently works very much like Google, except that instead of searching web pages it searches the traffic of the world's telephone and information network in real time.
Despite the fact that Echelon looks impressive (and frightening), several points should be taken into account. Firstly, the network built by the NSA and its allies was mainly aimed at the old USSR, where human intelligence was difficult, and the collection of enormous volumes of telephone calls gave at least some idea of what was happening. But the resulting political and economic intelligence turned out to be weak; the West thought that Russia's economy was about twice as large as it actually was, and was surprised by its collapse after 1989. (The incentives of the agencies to exaggerate the threat are obvious.) In any case, most of the effort was military, aimed at understanding Soviet radars and communications, as well as at obtaining a decisive advantage in location, suppression and deception. Without the ability to wage electronic warfare, a modern state cannot compete in air or naval battles or in tank engagements on the ground. Therefore it is not surprising that most of the NSA's personnel are military, and that its director has always been a serving general. A great deal of effort is still spent on understanding the signals of potential adversaries.
Secondly, there were some successes against terrorists — in particular, the arrest of the alleged organizer of 11 September, Khalid Sheikh Mohammed, after he had used a mobile telephone SIM card from a batch bought by a known terrorist in Switzerland. But the electronic war against the insurgents in Iraq turned out to be unproductive. And it has long been clear that far more effort should have been directed to human intelligence. In an article published shortly before 11 September, an analyst wrote: "The CIA probably does not have a single genuinely qualified Arabic-speaking officer of Middle Eastern origin who could plausibly play a Muslim fundamentalist and who would volunteer to spend years of his life with lousy food and no women in the mountains of Afghanistan. For God's sake, most case officers live in the suburbs of Virginia. We don't do those kinds of things". Another put it even more bluntly: "Operations that involve diarrhea as a way of life don't happen". The combination of remote technical intelligence and enormous firepower corresponds to the personal interests of bureaucrats and equipment suppliers, but it makes the allied forces ineffective in fighting insurgents, where the enemy mingles with the civilian population. Similar perverse incentives also hamper the military. For example, Britain spends billions on two new aircraft carriers and on modernizing its nuclear deterrent, but even six years after 11 September we have not trained enough soldiers capable of sustaining a basic conversation in Arabic. The big debates now brewing in the Pentagon concern not only intelligence, but also how to work out a smarter approach to counterinsurgency in general.
Thirdly, although other countries may complain about US signals intelligence collection, it is hypocritical of them to moralize about it. Other countries also conduct intelligence operations and are often far more aggressive in carrying out economic and other non-military espionage. The real difference between the WASP countries and the rest is that nobody else has built a "system of systems" like Echelon. Indeed, in the field of signals intelligence, as in others, network effects may operate: the value of a network grows faster than its size, and the more you intercept, the cheaper it becomes. Therefore attempts have been made to build a "European Echelon" involving the police and intelligence agencies of the continental European countries.
The mature point of view is that signals intelligence is necessary for the survival of the nation, but potentially dangerous — just like the armed forces it serves. The question is not whether such resources should exist, but how they are accountable. In the USA Senator Church's hearings in 1975 described in detail a number of abuses, such as the unlawful monitoring of US citizens. The collection of foreign intelligence is now regulated by US legislation in the form of 50 USC 1801–1811, which codifies FISA. This is not perfect; as already noted, it is the subject of bitter disputes between the executive and the legislature over the recent provision that the NSA may wiretap calls in the USA if one of the parties is believed not to be a US citizen. Even before this, the number of FISA warrants grew steadily after 11 September and exceeded the number of ordinary Title III wiretap warrants. But at least Congress has become interested. And the USA is lucky: in most countries oversight of intelligence is not even discussed.
Finally, poor accountability costs more than just the erosion of freedom and occasional political abuses. There is also a real operational cost in the proliferation of intelligence bureaucracies that turn out to be mostly useless when hostilities begin. In Washington during the Cold War the agencies hated each other far more than they hated the Russians. In Britain one of the most bitter intelligence battles was not against the IRA, but between the police and MI5 over who would take the lead in the fight against the IRA. There are many reports of intelligence inefficiency and internal squabbles from well-informed insiders such as R. V. Jones. It is in this context of bureaucratic wars over spheres of influence, which must be described, that [these events] became a formative experience for many governments (and NGOs) in questions of surveillance policy and technology.

The Crypto Wars

Technology policy in the 1990s was dominated by bitter debates about key escrow — the doctrine according to which anyone who encrypts data must provide the government with a copy of the key, so that civilian use of cryptography would not hinder the collection of intelligence by the NSA and others. Although some restrictions on cryptography had existed for years and irritated both academic researchers and civilian users, they hit the front pages in 1993, when President Clinton stunned the IT industry with the Escrowed Encryption Standard, better known as the Clipper chip. This was a proposed replacement for DES, with a built-in backdoor key so that government agencies could decrypt any traffic. The NSA had tried to sell this program to the cabinet of President Bush Sr. and had failed, but the new administration was happy to help.
Opinion in the USA became polarized. The government argued that, since cryptography is intended to preserve the secrecy of messages, it could be used by criminals to prevent the police from collecting evidence by wiretapping; the IT industry (with a few exceptions) took the opposite view, that cryptography was the only means of protecting electronic commerce and was thus vital for the future development of the network. Civil liberties groups sided with the industry and declared that cryptography would become a critical technology for privacy. By 1994 the NSA had concluded that it was facing a war with Microsoft, which Bill would win, so it handed the leadership of policy over to the FBI, continuing to direct questions from behind the scenes.
The debate quickly became intertwined with export control on weapons, by means of which cryptography had traditionally been controlled. American software firms were not allowed to export products containing cryptography that was too hard to break (this usually meant a key length of more than 40 bits). A software author from the USA, Phil Zimmermann, was called before a grand jury for arms trading after the program he had written — PGP — "escaped" onto the internet. He immediately became a folk hero and made a fortune, since his product captured the market leadership. The conflict became international: the US State Department tried to persuade other countries to control cryptography too. This became one of the personal missions of Vice President Gore (the reason why many people in Redmond and Silicon Valley contributed to Bush's campaign in 2000).
The results were mixed. Some countries with memories of oppressive regimes, such as Germany and Japan, resisted American persuasion. Others, such as Russia, took the opportunity to adopt harsh laws on the control of cryptography. France defiantly broke the traditional ban on non-governmental use of cryptography; Britain obediently moved from the liberal policy of non-intervention under John Major in the mid-1990s to a draconian law under Tony Blair. The Regulation of Investigatory Powers Act (RIP) of 2000 gives the police the right to demand that I hand over a key or password in my possession, and the Export Control Act of 2002 requires me to obtain an export license if I send any cryptographic software outside Europe using keys longer than 56 bits. Oh, and the government also obtained powers to vet foreign postgraduate students studying dangerous subjects such as computer science, and to refuse visas to those it considers a proliferation risk.

Did the Crypto Wars Matter?

When the debates about key escrow flared up in Britain in 1994–1995, a position was taken by many that was at the time unpopular both among the escrow lobbyists and among their opponents. The supporters of escrow argued that, since cryptography provides confidentiality and confidentiality can help criminals, there must be some way to get round it. The opponents of escrow argued that, since cryptography is necessary for confidentiality, there must be no way to get round it. Essentially all the premises of these arguments were wrong. Most applications of cryptography (in the real world, as opposed to the academic one) concern authentication, not confidentiality; they help the police rather than hinder them. As for criminals, they require inconspicuous communication — and encrypting a telephone call is a good way of attracting the attention of the agencies. As for confidentiality, most breaches occur as a result of the abuse of authorized access by insiders. Finally, a far more serious problem for policemen investigating electronic crimes is to find admissible evidence, for which good authentication can be useful.
This does not mean that the police have no need of wiretapping. Although many police forces manage quite successfully without it, and some of them, as we have seen, do use it, and the threat of wiretapping can be an effective deterrent, there are cases where it is necessary. The Walsh Report — by a senior Australian intelligence officer — gives a fairly balanced consideration of these questions. He compared the operational merits of wiretapping, bugs and physical surveillance and pointed out that wiretapping was either the cheapest or the only method of investigation in some circumstances.
Nevertheless, the intelligence community's main achievement in the 1990s, as was later discovered, lay in minimizing the number of systems that use cryptography by default. If a significant part of internet traffic had been encrypted, then the automated keyword searching performed by systems such as Echelon would have been largely thwarted. The NSA was perfectly aware that many new network systems were being built rapidly during the dot-com boom, and if cryptography had not been built into them from the very beginning, then most likely it would have been too expensive to introduce it later. Therefore every year the NSA held the line on the control of cryptography, which meant dozens of systems open to interception for decades to come. In these terms the policy was successful: only an insignificant part of the world's network traffic is encrypted, with some exceptions such as DRM-protected content, Skype, a few web pages protected by TLS, opportunistic TLS encryption between mail servers, SSH traffic, corporate VPNs and online computer games. Everything else is largely open to interception — including the masses of highly sensitive email between companies.

The Walsh Report and the European Dispute About Types of Keys

The Walsh Report — by a senior Australian intelligence officer — gave a balanced assessment of these problems. He concluded that there were no compelling reasons for the early regulation of cryptography, but recommended allowing the police and the intelligence services to hack the computers of suspects in order to obtain evidence (for example, by installing spyware on them).
In Europe these debates led to a split during the discussion of "Euroclipper". The German government demanded that only confidentiality keys be subject to escrow. Britain, however, insisted that digital signature keys be escrowed too. The British position rested on military doctrine, where deception and disinformation are considered no less important than wiretapping, while the Germans relied on police doctrine, which requires the preservation of the evidential base and the prevention of the forgery of signatures.

The Crypto AG Scandal and Hans Bühler

One of the most famous examples of covert control became the case of the Swiss company Crypto AG, which was the leading supplier of encryption equipment to governments that did not have their own technologies. In 1992 its salesman Hans Bühler was arrested in Iran. He stated that the company was covertly controlled by the intelligence services of the USA and Britain (GCHQ), and that backdoors had been built into its devices for years, known in professional circles as "red threads", which allowed the intelligence services to read the traffic.

Export Control

The main side effect of the crypto wars was the introduction of far more rigid export control than before, especially in Europe. Let us give a brief summary of the legislation on cryptography.
International arms control agreements (COCOM and the Wassenaar Arrangements) oblige most governments to introduce export control over cryptographic equipment, and the latter are implemented in the European Union by an EU regulation obliging member states to control and license the export of dual-use goods — goods that have both civilian and military application. Cryptanalytic products fall under the military regime, whereas the overwhelming majority of software that simply uses cryptography for protection falls under dual-use goods.
But national implementations differ. UK legislation did not control the export of intangible assets until 2002, so cryptographic software could be exported electronically; the Belgian government issues licenses for almost everything, and Switzerland remains a major exporter of cryptographic equipment.

Censorship

"The debates about encryption policy became in the 1990s a rehearsal for an even larger battle, which was fought around anonymity, censorship and copyright". And although the questions of copyright and DRM have largely stabilized, censorship has in recent years become a far more acute problem.
Censorship is carried out for the most varied motives. China blocks not only the sites of dissidents, but also emails containing forbidden words. Some countries introduce censorship during elections or after crises (as happened in Burma). In the USA there are debates about whether internet providers that are also telephone companies should block VOIP, and whether providers running cable channels should block P2P traffic (this is the question of the principle of net neutrality). In Europe the problem is that different countries ban different content: France and Germany prohibit the sale of Nazi paraphernalia. Many countries have tried to introduce control over child pornography — this has become the standard justification for politicians who want to "do something" about the Internet. There is also a European initiative to block radical Islamist sites. Finally, censorship is sometimes imposed by the courts in the context of civil disputes, for example the ban on publishing the DeCSS code.
The forms of censorship are also diverse: from the blocking of certain types of traffic and filtering by IP addresses to DNS poisoning, content inspection and out-of-band mechanisms (for example, punishing people who downloaded undesirable material).

Censorship by Authoritarian Regimes

China. At the end of 2006 there were 137 million internet users in China. The government invested enormous funds in filtering technologies. The "Great Chinese Firewall" is a complex socio-technical system providing layered defense against various materials: from pornography to religious materials and political dissent.

• Perimeter protection: Routers in Shenzhen filter by IP address the known "bad" sites (Voice of America, the BBC) and use DNS cache poisoning. Deep packet inspection at the TCP level is used to identify emails and web pages containing forbidden words (for example, "Falun Gong"): TCP reset packets are sent to both ends of the connection in order to break it. Keyword filtering (about 1000 forbidden words) is applied in Chinese search engines and blogs.
• Application-level protection: Some services are blocked and others are not, depending on how far the service provider plays by the rules of the regime. There were great disputes when Google agreed to censor its search results in China (in fact they fill their Chinese index using spiders that search from inside China, and therefore see only the material that is available there anyway). The incentives created by China's rapidly growing markets allow its government to make large international firms comply.
• Social protection: From 30,000 internet policemen to court cases against cyber-dissidents and laws requiring internet cafes to identify their customers, to a pair of cartoon mascots of the internet police (Jingjing and Chacha), who appear everywhere on the net to remind users that they are in a social space, and not a private one.

However, control is falling behind. There are more than 20 million blogs in China, and although the internet police energetically remove openly seditious material, online discussion of local news events has led to the appearance of a proper "public opinion" that for the first time is not in bondage to media managers. This is connected not only with email and blogs, but also with the rapid growth in the use of mobile telephones. Local events, such as the seizure of land by corrupt officials, can now rapidly rise up the agenda, subjecting the government to pressure from which it was formerly isolated.

Burma. The sudden rise in fuel prices in August 2007 led to mass protests and to a brutal suppression by the army from 26 September, in which perhaps several hundred people died. During the protests Burmese citizens used the internet and mobile communication to send photographs and videos to the outside world, with the result that their uprising made the front pages of the world's newspapers, and the suppression provoked wide condemnation of the junta. This happened despite the fact that Burma is one of only 30 countries in the world where less than 1% of the population uses the internet.
The world headlines clearly caused pain, and on 29 September a complete shutdown of the internet began. This was the first case where a mass blocking of the internet was used to prevent the spread of news. From 4 to 12 October a curfew was in force, when communication was available only from 22:00 to 4:00 in the morning; in the third stage some internet cafes were allowed to reopen, but the speed was limited to 256 Kbit/s, and others were closed and their equipment confiscated. Most Burmese used censorship-circumvention tools such as proxies. In fact the uprising was called the "g-lite revolution", after the popular Gmail proxy. If the lesson of this sad incident is that even 1% internet usage can destabilize a dictatorship, and that even dictatorships find it hard to do without the internet, then this is rather encouraging.

Net Neutrality

A number of less developed countries block VOIP services in order to facilitate the wiretapping of telephone conversations and to preserve the profitability of state telephone companies. Telephone companies in developing countries often receive a large part of their revenues from their share of the payments that foreigners make for calls into that country, and VOIP allows emigrants to avoid these payments.
However, most of the problems with VOIP arise in developed countries, especially in America. A number of ISPs are also telephone companies and use technical mechanisms to disrupt the operation of VOIP services — for example, by introducing jitter or short pauses into the packet flow. As a result, bitter debates about net neutrality flared up in Washington. On the one hand, the VOIP industry is in favor of a law that would oblige ISPs to treat all traffic equally; on the other hand, the telephone companies object: "Don't regulate the Internet".
The question is broader than just VOIP. Telephone companies have always charged completely different tariffs for different types of traffic: given that they have high fixed costs and low marginal costs, they have every incentive for price discrimination. Ed Whitacre, the chairman of AT&T, began these debates in 2005 when he stated that for companies such as Google, Yahoo or Vonage to use "his" broadband channels free of charge in order to make money for themselves was "insanity". This divided Congress mainly along party lines: the Democrats are in favor of net neutrality, and the Republicans in favor of the telephone companies.
In Europe net neutrality is less of a problem, since there is more competition in the ISP market there. Regulators tend to consider that if some ISPs engage in traffic shaping, this is not so important as long as customers can switch to other ISPs that do not.

Peer-to-Peer, Hate Speech and Child Pornography

The three "horsemen" promoted by the supporters of internet censorship in developed countries are file sharing, hate speech and child pornography.

File sharing. File sharing systems raise some questions of net neutrality; for example, Comcast disrupted the operation of BitTorrent using the same techniques of forging TCP reset packets as in the "Great Chinese Firewall". Comcast, being a cable operator, wants its customers to watch television over its cable channel rather than downloading it, in order to maximize its advertising revenues. The main players advocating the filtering of peer-to-peer traffic are the music companies. Many universities were intimidated by the threat of lawsuits and restricted such traffic in student local networks. Therefore it will be interesting to see whether countries introduce mandatory filtering for "moral" purposes, which the music industry will then be able to use for its own purposes as well.
Hate speech. In Europe there was a recent attempt to introduce an obligation for ISPs to filter hate speech, and in particular jihadist web sites. France and Germany prohibit the sale of Nazi paraphernalia. Experts are skeptical about whether such a law will make Europe safer; a ban on the writings of the radical Muslim Deobandi sect, to which perhaps a third of Britain's Muslims belong, will most likely worsen the tension in society. Moreover, research shows that most of the hate literature distributed inside and outside British mosques is produced or financed by religious institutions from Saudi Arabia. The government's reaction is not to confront King Abdullah, but to invite him on a state visit. Internet censorship here is a diversionary maneuver; it allows the government to state that it is doing something. It is better to leave this material in open access, as America does, and to let the police monitor the traffic on the worst of the sites, rather than forcing Muslim youth to acquire the skills of the Chinese and Burmese in using proxies.
Child pornography. In the 1990s, when governments were looking for some lever of influence over the Internet, the view formed that explicit images of sexual violence against children are about the only thing on which all states can agree that it should be banned. When they spoke in favor of a repressive measure — such as key escrow — they brought along people from children's charities who passionately asserted that the Stalinism of the day was vitally necessary to save children from harm. Needless to say, those of us who were on the liberal side of the argument would have preferred the charities to spend their money on campaigning about the more serious problems of child protection, such as the abuse of children in children's homes and child prostitution; and when a genuinely serious problem arose at the boundary between IT policy and child protection — the proposal to create a national database on child welfare, which would open the personal information of millions of children to hundreds of thousands of public sector workers — these worthy defenders of children were silent.
The debates about child pornography subsided in most countries (Russia is a notable exception; Putin uses child pornography as the main pretext for censorship aimed at political opponents), since terrorism took the place of child pornography as the ace up the executive's sleeve. But the hysteria did have evil consequences. They were severe in Britain, where it was used to justify not only more pervasive online surveillance, but also the creation of the National Hi-Tech Crime Unit. This unit carried out "Operation Ore", during which about eight thousand British citizens were subjected to police raids on suspicion of buying child pornography. It turned out that most of them had probably been the victims of credit card fraud. The pornography squad did not understand credit card fraud and did not want to; they were fixated on obtaining convictions for pornography. As a result, several thousand men suffered for months or even years because of a false arrest on extremely stigmatizing charges of which they were innocent. For some, the revelation that the police had messed up came too late; more than thirty men, faced with the prospect of a public prosecution that would probably destroy their families, committed suicide. The reason for all this was that the operators of illegal porn sites bought up lists of credit card numbers and then ran them through payment collection portals — probably believing that many people would not dare to report debit transactions for such services to the police. And although the police justified their operations by claiming that they would reduce harm to children, the suppliers of child pornography in the "Ore" case escaped prosecution.
America actually handled this case far better than Britain. About 300,000 US credit card numbers were found on the servers of the operator of the main portal, Thomas Reedy; the police used the names for intelligence, and not as evidence, matching the names against their databases, identifying suspects of interest — such as people who work with children — and quietly investigating them. More than a hundred convictions were obtained for actual abuse of children, and not a single false conviction. A pragmatic emphasis on old-fashioned police work is far preferable to the whipping up of fear and to grandiose political gestures.

Forensic Examination and the Rules of Admissibility of Evidence

This brings us naturally to the last main topic in the space of justice, namely how information can be extracted from computers, mobile telephones and other electronic devices for use as evidence. Three big changes in recent years: firstly, the sheer volume of data; secondly, the growth of search engines and other tools for finding relevant material; and thirdly, the fact that the courts are gradually becoming more relaxed and more competent.

Forensics

When the police carry out a raid even against a small-time drug dealer nowadays, they may obtain well over a terabyte of data: several laptops, half a dozen mobile telephones, a couple of iPods and perhaps a box of flash drives. The suspect may also have dozens of online accounts for web-mail services, social networking sites and other services. He may have interesting gadgets — for example, satellite navigation devices that store the history of his location (most of which is also available through the records of his mobile telephone). Security researchers have found a multitude of cunning ways of extracting information from data — for example, one can determine which camera took a picture from the noise pattern of the CCD array, and the number of such tricks can only increase.
The use of all this material as evidence depends, in most countries, on the observance of certain procedures. The material must be collected lawfully (with a search warrant or equivalent powers) and the forensic expert must maintain the chain of custody, which means being able to satisfy the court that the evidence has not been tampered with afterwards. The basic procedure consists in using tools that have been duly tested and evaluated to create reliable copies of the data (for example, computing a one-way hash of the data in order to establish their authenticity), documenting everything that is done, and having means for the proper handling of any private material (for example, privileged emails between a lawyer and a client).
Computer forensics poses ever more complex engineering problems. A recent example is that many police forces adopted a rigid procedure of always switching off PCs so that the hard disks could be mirrored. The Rockphish gang used this by making its phishing software memory-resident. The police would arrive at a house containing a phishing server, switch off the machine — and lose all the information that would have allowed them to trace the real server for which the seized machine was acting as a proxy.
A related problem is that Windows Vista comes with Bitlocker, a disk encryption utility that stores the keys in the TPM chip on the motherboard, and therefore makes the files unusable after the machine is switched off if you do not know the password. Although in Britain there is now a law allowing the courts to imprison people for refusing to hand over a password, in most countries there is not.
Another problem is that it may be important to minimize the disruption caused by forensic copying. Even if a machine is seized from a suspect, it may be problematic if you spend too much time examining it. In the cases of Operation Ore, many people who later turned out to be innocent had their PCs taken away and stored for months or even years, because the police had no forensic capacity to cope with it. As a result they remained under a cloud of suspicion far longer than was reasonable; this had an adverse effect on people in regulated professions, which led to lawsuits against the police. There is also the problem of loss of access to data, which for an individual or a small business can be catastrophic. Nowadays the prudent practice for a student is to have a backup protected from seizure, for example to set up a Gmail account and regularly send copies of his dissertation there.
Another trap of forensic examination is to rely on evidence extracted from the systems of one of the parties to a dispute without applying sufficient skepticism. Recall the Munden case, which was described in the earlier parts of Security Engineering. A man was falsely accused and falsely convicted of attempted fraud after he complained about unauthorized debits from his bank account. At the appeal hearing his defense team obtained a court order that the bank open its systems to a defense expert. The bank refused, the bank statements were declared inadmissible, and the case fell apart. Therefore, when relying on forensic evidence provided by a participant in a dispute, it is worth thinking in advance about whether it will withstand examination by hostile experts.

Admissibility of Evidence

When the courts first encountered computer evidence in the 1960s, there were many concerns about its reliability. There was not only the engineering question of whether the data were accurate, but also the legal question of whether computer data were inadmissible on the grounds that they are hearsay. Various legislatures resolved this in different ways. In the USA most of the law is contained in the Federal Rules of Evidence, where computer records are usually introduced as business records. In the UK the situation is similar: the Civil Evidence Act 1995 covers civil proceedings, while the Police and Criminal Evidence Act 1984 concerns criminal cases. The requirement that a machine be operating in the ordinary course of business operations can cause problems when machines have to be operated in abnormal ways in order to extract information.
There exist some special legal provisions for certain technologies; many of them were adopted during or shortly after the dot-com boom, as legislators strove to pave the way for electronic commerce without understanding the problems. Many industry lobbyists argued that electronic commerce was being hindered by uncertainty about whether electronic documents would be accepted as a "written form" for those laws that require certain transactions to be in writing. Therefore legislatures, starting with Utah, adopted laws granting special status to digital signatures. In most cases this had no effect at all, since the courts adopted the sensible opinion that an email is a written form just as much as a paper letter: the essence of a signature is the intention of the signer, and the courts have long decided in this way. For example, an agricultural worker who was crushed to death by a tractor and managed to scrawl "all to mum" on a tire was held to have made a legal will.
However, there is one case where impatient legislators got it completely wrong, and that is Europe. The Electronic Signatures Directive, which came into force in 2000, obliges all member states to give special force to the advanced electronic signature, which basically means a digital signature generated with the help of a smart card. Europe's smart card industry thought that this would bring them a lot of money. However, it had the opposite effect. At present the risk that a paper cheque will be forged is borne by the party relying on it: if someone forges a cheque on my account, then it is not my signature and I did not give the bank my mandate to debit my account, so if they negligently rely on a forged signature and do so, that is their problem. However, if I were foolish enough ever to accept an advanced electronic signature device, then a presumption of validity would arise for any signature that appears to have been made with its help. Suddenly the risk shifts from the bank to me. I become liable to anyone in the world for any signature that appears to have been made by my smart card, regardless of whether I actually made it! This, together with the facts that smart cards have no reliable user interface, and that the PCs which most people would use to provide that interface are easily and often hacked, made electronic signatures instantly unattractive.
Finally, a word about click-wrap. In 2000 the US Congress passed the Electronic Signatures in Global and National Commerce Act (ESIGN), which gives legal force to any "sound, symbol or process" by which a consumer agrees to something. Thus, by pressing a key on a telephone, clicking a hyperlink to enter a web site or clicking "continue" in a software installer, the consumer agrees to be bound by a contract. This makes click-wrap licenses work in America. The general opinion of lawyers in Europe is that they probably do not work here, but nobody is in a hurry to bring the first case.

Confidentiality and Data Protection

Data protection is the term used in Europe to denote the protection of personal information from improper use. Personal information usually means any data stored about an identifiable person or data subject, such as bank account details and credit card purchase patterns. This corresponds approximately to the American term computer privacy. The difference in terminology is accompanied by an enormous difference in laws and attitudes.
European legislation gives data subjects the right to inspect the personal data stored about them, to alter them if they are inaccurate, to understand how they are processed, and in many cases to prevent their transfer to other organizations without their consent. There exist exceptions for national security, but they are not as complete as the intelligence services would like: there was a big scandal when it emerged that data from SWIFT, which processes interbank payments, were being copied to the Department of Homeland Security without the knowledge of the data subjects. The European data protection authorities ruled that SWIFT had violated European, Belgian and Swiss data protection legislation, and it agreed to cease processing European data in the USA by the end of 2009.
Almost all commercial data are covered, and there exist particularly strict rules for controlling data concerning intimate matters such as health, religion, race, sexual life and political beliefs. Finally, a recent law prescribes that personal data may not be sent to organizations in countries whose laws do not provide comparable protection. In practice this means America and India, where the legal protection of privacy is fragmentary. The solution so far has been the safe harbor agreement, under which a data processor in America or India promises its European customer to comply with European legislation. Many firms do this; the pioneer was Citibank, which created such an arrangement for processing the data of German cardholders in South Dakota. But this creates practical problems of enforcement for EU citizens who consider that their rights have been violated; they are not a party to the contract and may find it hard to persuade the US Department of Commerce to take action against a US firm. Therefore the safe harbor provisions may fail when tested in court.

European Data Protection

Technophobia is not an invention of the end of the twentieth century. Back in 1890 the judges Warren and Brandeis warned of the threat to privacy from "recent inventions and business methods" — in particular, photography and investigative journalism. Years later, after large retail companies began to use computers in the 1950s and banks followed them in the early 1960s, people began to worry about the social consequences if all the transactions of a citizen could be collected, consolidated and analyzed. In Europe big business avoided condemnation by putting forward the argument that only the government could afford enough computers to represent a serious threat to privacy. As soon as people realized that it was economically and rationally expedient for the government to extend its power by using the personal data of all citizens as the basis for prediction, this became a question of human rights — given the recent memory of the Gestapo in most European countries.
A patchwork of data protection laws began to appear, starting with the German state of Hesse in 1969. Because of the speed at which technologies change, the successful laws were technologically neutral. Their common theme was a regulator (at national or state level) to whom users of personal data had to report and who could order them to cease and desist from improper processing.
Over time the data processing of transnational companies also became a problem, and people realized that purely local or national initiatives would probably be ineffective against them. After a voluntary code of conduct adopted by the OECD in 1980, data protection was enshrined by a Council of Europe convention in January 1981, which came into force in October 1985. Although strictly speaking this convention was voluntary, many states signed up to it for fear of losing access to data processing markets. It required the signatory states to adopt domestic legislation implementing at least some minimum guarantees. Data were to be obtained lawfully and processed fairly, and the states had to guarantee that legal remedies were available when violations arose.
The quality of implementation varied greatly. In Britain, for example, Margaret Thatcher shamelessly did the minimum possible to comply with European legislation; a data protection body was created, but it was deprived of funding and technical expertise, and many exceptions were granted for selected groups of voters. In the strict privacy countries, such as Germany, the data protection authorities became serious law enforcement agencies.
By the beginning of the 1990s it had become clear that the difference between national laws created barriers to trade. Many companies avoided control altogether by moving their data processing to the USA. Therefore data protection was finally raised to the rank of full-fledged European legislation in 1995 with the Data Protection Directive. It sets higher minimum standards than most countries had required before this, with particularly strict control over highly sensitive data such as health, religion, race and political affiliation. It also prevents personal information from being sent to "data havens" such as the USA, unless comparable controls ensured by contract operate there.

The Differences Between Europe and the USA

The history in the USA is described in detail in my articles on security; on the whole, business managed to persuade the government to leave privacy largely to "self-regulation". Although there is a patchwork of state and federal laws, they are application-specific and heavily fragmented. On the whole, privacy in federal government records and in communications is regulated fairly strictly, while business data are largely uncontrolled. There are a few islands of regulation, such as the Fair Credit Reporting Act of 1970, which regulates the disclosure of credit information and is generally similar to the European rules; the Video Privacy Protection Act or "Bork Bill", adopted after a Washington newspaper published the story of the video rentals of Judge Robert Bork after his nomination to the US Supreme Court; the Drivers' Privacy Protection Act, adopted to protect the privacy of DMV records after the actress Rebecca Schaeffer was killed by an obsessed fan who hired a private detective to find her address; and the Health Insurance Portability and Accountability Act, which protects medical records. However, US privacy legislation also includes several torts that provide the basis for a civil suit, and they cover a surprising number of circumstances. There was also a landmark case in 2006, when Choicepoint paid 10 million dollars to settle a suit brought by the FTC after it had failed to check subscribers properly and had allowed fraudsters to buy the personal information of more than 160,000 Americans, which led to at least 800 cases of "identity theft".
Attitudes also differ. Some researchers report a growing feeling in the USA that people have lost control over the use of their personal information, while in some European countries privacy is regarded as a fundamental human right requiring vigorous legislative support; in Germany it is enshrined in the constitution. But it must be said that there is a constant problem here. People say that they value privacy, but act differently. The overwhelming majority of people, whether in the USA or in Europe, will exchange their privacy for very small benefits. Privacy-enhancing technologies were offered for sale, but most of them failed in the market.
Nevertheless, in other countries the courts became more protective of citizens' rights after 11 September. In Germany, which usually takes the strictest position, privacy outranks even the "war on terror": the highest court declared unconstitutional the police action of 2001 to create a dossier on more than 30,000 male students or former students aged between 18 and 40 from Muslim countries — even though nobody was arrested as a result. It ruled that such exercises may be carried out only in response to specific threats, and not as a preventive measure.
The other side of the coin of privacy law is freedom of information law. A radical version of this is offered by David Brin. He reasons that the falling cost of acquiring, transmitting and storing data will make ubiquitous surveillance technologies available to the authorities, so the only real question is whether they are available to the rest of us as well. He paints a choice between two futures — one in which citizens live in fear of an East-German-style police state, and one in which officials are accountable to public scrutiny. Cameras will exist: will they be surveillance cameras or webcams? He argues that essentially all information should be open — including, for example, all our bank accounts. Weaker versions of this have been tried: tax returns are published in Iceland and in some Swiss cantons, and this practice reduces evasion, since rich people fear the loss of social status that an artificially understated declared income would bring.
There are some interesting engineering questions. For example, while convictions for serious crimes in the USA remain on the records forever, in many European countries there are laws on the rehabilitation of offenders under which most convictions disappear after a period of time that depends on the gravity of the crime. But how can such laws be enforced now that search engines exist on the Internet? The German reaction is that if you want to cite a criminal case, you must obtain an officially anonymized transcript from the court. In Italy a convicted businessman obtained a court decision to remove from a government web site the record of his conviction, after the term of the conviction had expired. But if the electronic archives of newspapers are searchable on the Internet, what use is this — unless the names of all offenders are blocked from electronic reports? Recently there have been many debates about the monitoring of former sex offenders who committed crimes against children, with laws in some states requiring offender registers to be publicly available, and riots in Britain after a Sunday newspaper named some former offenders. How can offenders be rehabilitated in a world with Google? For example, do you mark the names of offenders in newspaper reports of trials with an expiry date and pass laws obliging search and archive services to observe them?
The point is that even if data are publicly available, their use may still cause offenses under European data protection legislation. This creates peculiar difficulties in the USA, where the courts have consistently interpreted the First Amendment as meaning that you cannot stop the repetition of true statements in peacetime, except in a small number of cases.
Therefore it is not surprising that the current point of conflict between Europe and America on questions of privacy concerns Google. The immediate casus belli is that EU legislation requires personal data to be deleted as soon as they are no longer needed, while Google built its systems to store such data as click streams indefinitely. During 2007 the European data protection authorities brought this to Google's attention; the search engine offered to anonymize click streams after 18 months. Given the complexity of properly carrying out control over the outputs, the statement will certainly be closely examined by the European authorities. There is no doubt that this saga will continue. Even in America there was a call from the CDT and the EFF for a "Do Not Track" list, analogous to the "Do Not Call" list, so that people could opt out; other activists disagree, saying that this would undermine the model of useful web services paid for by advertising. In any case, less than one percent of people use ad-blocking software. We shall have to wait and see.

Conclusions

Governments and public policy are increasingly intertwined with the work of the security engineer. The "crypto wars" were a harbinger of this, as was the struggle over copyright, DRM and Trusted Computing. Current problems also include surveillance, confidentiality, DRM, consumer protection and even electronic voting, which are acquiring comparable importance.
The biggest technical challenge will probably be connected with systems integration and assurance. Ten years ago the inhabitants of the various islands in the security archipelago all had enormous confidence in their products. Cryptologists believed that certain ciphers could not be broken, smart card suppliers asserted that the extraction of the cryptographic keys stored in their chips was absolutely physically impossible, and the security printing people said that holograms could not be forged without a PhD in physics and $20 million worth of equipment. At the systemic level too there was a lot of misplaced confidence. Banks asserted that their ATMs could not even in theory make an erroneous debit, the multilevel security crowd sold their approach as the solution to all problems of system protection, and people assumed that a security evaluation carried out by a laboratory licensed by the government of a developed country would be both honest and competent. All these comfortable old certainties have evaporated. Instead, security has become part of the broader problem of dependability. We build ever better tools, and they help system builders to climb a little further up the mountain of complexity, but in the end they fall off. The proportion of large complex systems projects that fail is just as it was in the 1970s; but now we build far bigger catastrophes.
Complexity is the real enemy of security. The distinction between outsiders and insiders used to simplify the business, but as everything becomes connected it is rapidly disappearing. Protection used to be based on a few big ideas and on propositions that could be formulated precisely, whereas now the subject is far more diverse and includes a lot of imprecise and heuristic knowledge. The system life cycle is also changing: in the old days a closed system was developed in a finite project, whereas now systems evolve and accumulate functions without limits.
The changes in the nature of work are significant: whereas formerly the chief internal auditor of a bank would remember all the frauds of the last thirty years and would not allow the data processing department to repeat the mistakes that caused them, the new corporate culture of temporary employment and "perpetual revolution" (as Mao described it) has destroyed corporate memory. Economics will continue to ensure that insecure systems are built — and responsibility will be dumped onto others where possible. Governments will try to keep up, but they are too slow, and they can often be bribed for a time. Therefore there will be plenty of regulatory failures too.
The net effect of all these changes is that the protection of information in computer systems is no longer a scientific discipline, but an engineering one. The security engineer of the twenty-first century will be responsible for systems that constantly evolve and face a changing spectrum of threats. He will have a large and constantly growing set of tools. A significant part of his work will consist in keeping up technically: understanding the latest attacks, learning to use new tools and keeping abreast of the legal and political fronts. Like any engineer, he will need a solid intellectual foundation; he will have to understand the basic disciplines such as cryptology, access control, information flow, networks and signal detection. He will also need to understand the basics of management: how accounts work, the principles of finance and his client's business processes. But most important of all will be the ability to manage technologies and to play an effective role in the process of the system's evolution in order to satisfy changing business needs. The ability to communicate with business people, and not only with other engineers, will be vital, and experience will be of enormous importance. I do not think that anyone with this combination of skills risks being left without work — or being bored — in the near future.
Finally, the unbridled growth of the security military-industrial complex after 11 September and the blatant whipping up of fear by many governments is a scar on the world and on our profession. We have a duty to help it heal, and we can do this in many different ways. In the world there is a development of the disciplines of security engineering and security economics, so that we can say not only what works in the laboratory, but also what can be made to work in the world. The dissemination of knowledge is also important — which is what I am doing with these articles on the subject. Economic growth also helps, and education: it is precisely the poor and the uneducated who are most susceptible to the whipping up of fear (whether from Western leaders or from bin Laden). At least in countries with an educated population, voters have begun to recognize the excesses of the "War on Terror" and to cope with them. Just as individual people learn from experience to compensate for our psychological biases and to cope with risk more rationally, so our societies learn and adapt too. Democracy is the key mechanism for this. Therefore the last way in which security engineers can make their contribution is participation in political debate. The more we can involve the people who lead the discussions about emerging threats, the faster our societies will adapt to cope with them.