Exploring the wreck of the OCSEC-2017

The Office of the CSE Commissioner, CSE's soon to be replaced watchdog agency, released its 2016-17 Annual Report back in August 2017. As is traditional, it almost immediately sank from sight and was lost to all human ken. Nearly a year later, I guess it's about time I mounted my annual expedition to see if there's anything worth salvaging from it. With luck, I might manage to raise a few items before the 2017-18 report is launched.

Unlike the 2015-16 report, this report did receive a modicum of media coverage in the immediate wake of its release, specifically on the issue of information-sharing with allies. (See Justin Ling, "Canada still hasn’t developed new rules for intelligence sharing with U.S. and allies," Vice News, 24 August 2017 and Alison Crawford, "Canada's electronic spy agency to get new rules for sharing data with allies," CBC News, 29 August 2017.) But I'm willing to bet there's still lots of material worth examining lying in the forgotten hulk.

So let's get this expedition underway.


Use/retention of private communications up 25,653%

OK, here's something interesting. According to the CSE Commissioner, in 2015-16 CSE used or retained 3,348 "private communications" that were collected under the agency's foreign intelligence program (see page 39 of the report).

In Canadian law, a private communication is a communication with at least one end in Canada. CSE's foreign intelligence program is not allowed to "target" Canadians or any person located in Canada, but if a foreign target of the agency who is located outside Canada communicates with someone inside Canada, CSE is permitted to collect that private communication as long as there is a Ministerial Authorization permitting such collection in place (and, rest assured, there is). The 3,348 figure reported by the Commissioner represents only one portion of the total number of private communications collected or otherwise acquired by CSE under the three parts of its mandate, but it's a potentially important indicator of how often Canadians get pulled into CSE's foreign intelligence collection activities.

I've been using highway signs to depict the private communications numbers reported by the Commissioner. In 2012-13 the number was 66 and in 2013-14 it was 17, later revised to 13. Last year it was 342, which was a bit of a challenge but I did find a suitable highway. This year I've had to improvise...



That's a big number. The Commissioner's report comments that the 2015-16 total is "almost 3,000" higher than the previous year total, which seems like an unusual way to put it since the actual difference is 3,006. Maybe the 2014-15 number was revised too. In any case, the two numbers aren't strictly comparable, as the 2014-15 figure refers to a seven-month period, while the 2015-16 figure covers a full twelve months. To get an apples-to-apples comparison, we need to go back two years to the 13 private communications used or retained over the twelve months of 2013-14.

Those figures show that the number of private communications used or retained by CSE's foreign intelligence program jumped by 25,653% between 2013-14 and 2015-16. That's a comma, not a decimal point: Twenty-five thousand six hundred and fifty-three percent.

So, yeah. Quite a big jump.

We do get an explanation of sorts for the change: "The increase in the number of used or retained private communications remains a consequence of the technical characteristics of certain communications technologies, and CSE’s legal obligations to count private communications in a certain manner."

But that doesn't really answer many questions.

In 2016, when this growth trend first became apparent, I speculated that CSE may be collecting an increasing number of communications transmitted by chat applications such as Facebook Messenger. Because each individual comment in such conversations is a separate transmission, it is likely that each would be considered a separate private communication for legal purposes. Thus, a single conversation lasting a just few minutes might contain dozens of private communications. If this is what explains the dramatic jump in the numbers since 2013-14, there may have been little if any actual increase in the number of persons in Canada whose conversations or other communications are being caught in CSE's dragnet.

That would certainly explain the Commissioner's apparent lack of concern about the numbers.

The current report doesn't confirm that theory (or provide any other intelligible explanation), but it does comment that "the current manner in which CSE counts private communications provides a distorted view of the number of Canadians or persons in Canada that are involved in (i.e., are the other end of) CSE interceptions to obtain foreign intelligence under ministerial authorizations."

And the report provides one additional key piece of information: The 533 private communications that were actually used in CSE's foreign intelligence reporting in 2015-16 (as opposed to temporarily retained for possible future use) appeared in a total of just 20 end product reports. This means that on average 26.65 private communications were cited in each one of those reports. Since some reports almost certainly concerned just a single private communication, many of them are likely to have cited 40 or 50 or more.

A little background on SIGINT end product reports might be helpful here. CSE does not produce extended intelligence assessments — it reports SIGINT facts, such as a single key piece of information overheard in an intercepted phone call. CSE analysts don't sit on such intelligence: they disseminate it to their clients in an individual end product report with as little delay as possible. If 20 or 30 or 40 private communications appear in a single end product report, it is because all of those communications were acquired at essentially the same time. And if this is happening routinely, it's almost certainly because the communications systems that CSE has begun to frequently target routinely generate large numbers of private communications at a time.

Which sounds like chat apps to me.

If these numbers do indicate growing collection of chat-related traffic by CSE, it would appear that the increasing use of encryption in those apps has not had the effect of shutting CSE out of that traffic — at least, not as of 2015-16. Are CSE's targets using insecure messaging apps, or versions that have been "enabled" to undermine their security? Are end-point operations, such as implanting malware on target smartphones, being used to bypass encryption? Given the high level of concern expressed by intelligence and security agencies in recent years about the prospects of "going dark", it will be interesting to see if the number of private communications used by CSE drops off in future reports.

I suspect CSE won't be entirely pleased to see this kind of speculation bandied about — even if my specific guesses are completely off base, which they may well be — so let me just suggest to the agency that if you were instead to declassify figures such as the number of individual persons in Canada who appeared in end product reports that year, the number whose identity information was released to clients at least once, and the total number of reports in which private communications were cited, the public would get figures much better suited to monitoring the privacy implications of CSE's operations, those figures would probably be more reassuring than the ones we get now (and if they're not, all the more reason to release them), and CSE's targets would be denied any basis for speculating as to the types of communications being monitored.

On page 4 of his report, the CSE Commissioner makes a direct plea for greater openness by CSE, highlighting "the need to re-examine what information is able to be disclosed to the public in an effort to promote transparency. Transparency has been a cornerstone of my approach as Commissioner. There have been significant strides in this regard in the United Kingdom and in the United States. It is time to do likewise in Canada."

Seems like a good idea to me.


More to come on the report in future posts (I hope).

Update 9 July 2018:

Stage two here.