The Disingenuous Dr. Halderman

Until a week ago Professor J. Alex Halderman of the University of Michigan was pretty much unknown outside the world of cyber-security, in which he holds a place of distinction as an academic expert on security problems concerning elections in the U.S. and other countries. But this technical expert has put himself on the way to becoming a household name since he reportedly advised the Clinton campaign team that they should request recounts of the presidential balloting in Wisconsin, Michigan, and Pennsylvania because of suspicion that Russia had hacked into their computerized voting systems and affected the outcomes. Jill Stein’s petition for a recount of the Wisconsin balloting includes an affidavit from Professor Halderman, which Stein cites as the sole “basis of belief” that irregularities occurred in Wisconsin that support her petition for a recount. So what does Professor Halderman say in his affidavit?

The one point on which Professor Halderman lays the greatest stress is that he, Professor J. Alex Halderman, is a really, really big expert in cyber-security. Exhibit A in the affidavit is Halderman’s CV, which is long and impressive (especially since he only received his Ph.D. in 2009).

The second point that Halderman stresses is that during 2016 there were various breaches of cyber-security, such as the breach of the DNC e-mail archive, that were apparently intended to affect the U.S. presidential election. Reports of these breaches constitute Exhibits B-F of the affidavit. It should be noted that only Exhibit D and Exhibit F concern alleged or suspected breaches of actual voting equipment. Exhibit D concerns a breach of the Illinois voter registration database that resulted in the theft of voter information data, and an attempted, but apparently unsuccessful, breach of the voter registration database in Arizona. (Please note that Wisconsin, where the recount is being sought, was not one of the states reportedly breached, and that neither Illinois nor Arizona were among the states Halderman reportedly recommended for recounts.) Exhibit F concerns an anonymously sourced report from DHS about attempts, apparently unsuccessful, to breach voting systems in as many as 20 unnamed states.

Halderman’s third point is that the very nature of many electronic voting systems renders them susceptible to compromise, not by internet hackers, but by the insertion of malicious software on-site. This part of the affidavit is not accompanied by evidentiary exhibits.

Halderman’s fourth point concerns speculation about how a foreign power intent upon affecting a U.S. presidential election would go about it. According to Halderman, after conducting advance espionage on election equipment at many actual voting sites, the foreign power would then rely on polling data to determine which states would have close electoral margins, and then spread malware into voting machines in some of those states. This point in Halderman’s affidavit is also not supported by any evidentiary exhibits.

Now this omission is peculiar, because state polling data from the days and weeks before the election is abundantly available, and it could easily be shown that certain specific states were indeed expected to be very close. So why, after overwhelming the Wisconsin Elections Commission with evidentiary exhibits that have no direct relevance to the process of casting, recording, and tabulating votes, did Halderman fail to provide evidence to support the plausibility of his speculation about what the hypothetical cyber-attackers actually might have done to affect the election?  

Well, this uncharacteristic lapse of thoroughness creates an obscurity in Halderman’s affidavit that is convenient for his case, because the states that the polls predicted would be very close do not include Wisconsin or the other two states where Halderman has recommended that recounts should be sought. On Monday, November 7, four polls in Florida (29 electoral votes) averaged to an exact tie (all data comes from the RCP site). Three November 6 polls from Ohio (18) gave Trump an average advantage of .3%. Three polls in Nevada averaged to an advantage of .7% for Clinton. Three polls in North Carolina (15) averaged to a one-point advantage for Clinton. Professor Halderman has not suggested that any of these four states (of which the closest went to Clinton) should be investigated for possible hacking, although according to his scenario they would have been among the most likely targets.

Meanwhile, in Wisconsin, for which Halderman has advised a recount on the grounds that it was a likely target for hacking, two polls published on November 5-6 gave Clinton an average advantage of 7%, suggesting that Clinton had the state safely in the bag. In Michigan, three polls published on November 6-7 averaged to an advantage of 2.7% for Clinton (two of them had Clinton ahead by 5%). On November 5 two polls in Pennsylvania gave Clinton an average advantage of 3%. Therefore, a hypothetical saboteur would not have concluded, on the basis of pre-election polls, that Wisconsin, Michigan, and Pennsylvania would be among the most closely decided states. (And as it turned out, the most closely decided state, by far, was New Hampshire, which Halderman also did not recommend for a recount. I wonder why.)

I have no reason to doubt that Professor Halderman possesses expert knowledge of computers and the technical means of compromising their security. His expertise in how to actually affect a national 50-state election is much less clear, and not supported by any evidence he presents. Perhaps most importantly, Professor Halderman’s speculative and conveniently omissive affidavit gives me the impression that his expert advice is not rendered in good faith. Maybe there is a job for him at MIT alongside the distinguished Professor Jonathan Gruber. 

If you experience technical problems, please write to