In justifying its bulk surveillance activities, the State Security Agency (SSA) has resorted to well-rehearsed arguments that spy agencies have made elsewhere. They have argued that bulk surveillance is an internationally accepted method of monitoring transnational signals to screen them for certain key phrases and to ensure that the country is protected from threats from individuals or organisations outside South Africa’s borders.
The SSA says that bulk surveillance is an automated process, with no real human intervention. So the traffic that is selected for further analysis is selected by machines, not people, and the rest is discarded without anyone having looked at it.
The SSA has also argued that it has a vested interest in retaining only that data that is of interest to them, because it would be too expensive to store too much. It has also set out its understanding of what constitutes foreign signals intelligence, which “… includes any information that emanates from outside the borders of the republic (in this case, South Africa) and passes through or ends in the republic”.
Conceivably, this definition could include the communications of locals as well, because some of their internet traffic is likely to be routed through foreign-based servers (particularly in the United States), and could qualify as a foreign signal if a South African is pulling data from a foreign server.
Britain’s Government Communications Headquarters (GCHQ) has recognised this complexity but has muddied the waters even further by arguing that a social media network such as Facebook is a “platform” as opposed to a service.
Therefore, in practical terms, for GCHQ, communications such as commenting on a friend’s post would qualify as an external signal, even if all the friends are in Britain. This assertion is based on the misapprehension that friends communicate not with one another in the first instance, but with the platform. Although this may not be the correct legal interpretation of foreign signals, it is possible that the SSA interprets the definition in the same way that GCHQ does for operational purposes.
The SSA infers that the privacy effects of bulk surveillance are minimal to nonexistent, because people are not intervening in the process. Furthermore, they argue that surveillance is not arbitrary: it is informed by Cabinet’s national intelligence priorities, and includes “classical” national security concerns (such as terrorism and organised crime), as well as national security concerns informed by South Africa’s human security mandate (such as food and water security, and illicit financial flows).
The SSA also has a vested economic interest in not storing too much data, because the costs of doing so would be prohibitive, so they are forced to be discerning in what they store. There are other controls that the SSA claims exist, such as the oversight provided by the joint standing committee on intelligence in terms of the 1994 trio of intelligence Acts, as well as the inspector general of intelligence.
Furthermore, the SSA has claimed that its bulk surveillance capacities are covered by the Regulation of Interception of Communications and Provision of Communication-Related Information Act, in that the Act authorises the establishment of interception centres, and bulk surveillance falls within the definitions of various forms of communications in the Act. Furthermore, any more information about the National Communications Centre and its activities could expose operational methods, which the SSA considered to be detrimental to national security.
But the SSA’s arguments do not meet the six-part test laid down for strategic intelligence-gathering that is recognised by the European Court of Human Rights; a test that any mass surveillance law should meet.
Although the ruling predated the kinds of internet-based mass surveillance that is being discussed here, the case did concern strategic surveillance of telecommunications networks to avert national security threats.
The test includes a clear statement on the nature of the offences giving rise to the surveillance; clearly defined categories of people liable to have their communications intercepted; limits on the duration of interception; adequate procedures for the examination, analysis and storage of the data obtained; adequate precautions when disseminating data to other parties; and clear statements about when the data obtained may be erased or the records destroyed.
Although the government has the discretion to decide when to engage in strategic surveillance, this cannot be an unfettered discretion and the public needs to know the circumstances that will trigger such surveillance.
Compounding the problem is the fact that there is no provision for reasonable suspicion that interception subjects have committed crimes, there is no prior independent authorisation of mass surveillance by a judge, even through bulk warrants, and also there is no requirement for subsequent notification of interception subjects of the interception measures taken against them.
The selection of the data of interest required human intervention in the communications flow; this human intervention happened when the SSA decided on the selectors to be used to sift the data of interest. So privacy was affected at that point.
In fact, privacy violations began when there was an interference with the ordinary flow of data, when it was diverted from its intended paths into the country. Selectors are applied only once the traffic has been extracted, which is where the initial intrusion occurs.
Furthermore, the SSA’s argument that it lacks the resources to store large amounts of data is cold comfort and does not address the issues of principle. If the resources available to the SSA increased, or its technical ability to store more data improved, then the inference of its argument is that it would collect and store more data. As a result, the resources argument cannot be considered to be a safeguard.
Deciding on selectors is a subjective process, which can lead to analysts’ assumptions and even prejudices being brought to bear on the sifting process. The broader the range of selectors, the more subjective the process is likely to be.
The SSA has made it clear that its grounds for monitoring the electromagnetic spectrum are broad, which may not be supported by the narrowed definition of national security in the General Intelligence Laws Amendment Act.
Also, the broad grounds for monitoring the spectrum imply soft as opposed to hard selectors: the softer a selector, the greater the potential for privacy violations because these selectors will intercept more traffic than hard selectors.
South Africa does not have a foreign signals intelligence-gathering court, so the selectors used are not approved by a court process, which implies that their approval is undertaken internally, and which increases the potential for abuse.
Furthermore, the SSA has not provided a clear statement about the offences that trigger bulk surveillance. The application of the human security definition of national security to this process makes this requirement even more necessary.
There is also no requirement for individualised reasonable suspicion, which means that the communications of innocents are being diverted and stored, with no limits on the duration of the surveillance or the categories of people to be surveilled.
Whether or not the use of mass surveillance is proportional to the level of external threat the country faces is an important consideration, because the SSA has claimed repeatedly that it faces no major external threat to national security.
Professor Jane Duncan is head of the department of journalism, film and television at the University of Johannesburg. This is an edited excerpt from her book Stopping the Spies: Constructing and Resisting the Surveillance State in South Africa (Wits University Press)