Prosecutors won't be able to use some Apple phone data against a man charged with murdering a Dorchester man then setting his building on fire, court rules

Suffolk County prosecutors who plan to use cell-tower data to place a suspect in a 2015 Dorchester home invasion, murder, shooting and arson at the scene of the crime won't be able to use data from the man's phone that seems to also place him at the location, the Supreme Judicial Court ruled today.

Although prosecutors will be allowed to use "cell site location information" from the cell towers pinged by the Apple phone they seized from Victor Arrington that they say puts him at the scene, the state's highest court said they cannot also introduce Apple iOS "frequent location history" (FLH) or "significant history" data an analyst in the Suffolk County DA's office says he was able to extract from the phone.

The justices said no court in the state - or anywhere else in the country, as far as they could tell - has allowed the use of FLH data, which rests on an encrypted Apple algorithm to generate a sort of index of where a user has been in recent days, and for which nobody has yet made the case that it is reliable enough to be introduced as evidence in a criminal trial.

This, the court said, is unlike cell-site location data, which can place a phone in the general vicinity of a specific location based on its recorded connections with nearby phone-company antennas and for which there is a relatively long history of both technical articles proving its accuracy and court decisions allowing its use.

Prosecutors had argued the data was critical to supporting their case that Arrington and two accomplices allegedly burst into an apartment at 332 Harvard St. in Dorchester on March 15, 2015, bound Richard Long and a woman, shot both - with the woman surviving the gunshot wound to the head - and then set the building on fire to try to cover their tracks.

The court acknowledged that:

Because its case at trial will rely significantly on testimony by a cooperating witness procured in exchange for a reduced sentence, we agree with the Commonwealth that evidence corroborating the cooperating witness's testimony (and, thus, bolstering the credibility of that testimony) might fairly be described as "critical" to the prosecution's case.

But the court said prosecutors have other evidence on which to try Arrington, including that witness's testimony, the phone location data and surveillance videos, so it's not like they now have to just drop the case.

Allowing the "frequent location" data would be wrong for a variety of reasons, under a legal doctrine that requires extensive proof of validity before information from a new technology can be admitted as evidence, and prosecutors failed to provide that, the justices concluded.

They noted the analyst who studied it acknowledged he used a hacked Apple phone that uses a different version of iOS than the phone seized from Arrington, that he was unable to take a look at the unencrypted algorithm Apple used and that the DA's office did not introduce any literature from technical journals that using data extracted from Apple FLH was accurate. Also, just because the data which the Apple software used to generate its "frequent location history" might be accurate, that doesn't mean the resulting FLH data was. And:

The trial judge found that the analyst's experiments with FLH data had a small sample size. To test the reliability of FLH data, the analyst jailbroke an iPhone 5C, which allowed him access to information that is ordinarily encrypted and inaccessible to an iPhone user. He brought the jailbroken iPhone to five different locations, examined the underlying location data points gathered by the iPhone at those locations from sources such as CSLI, Wi-Fi, and GPS, and compared these underlying location data points to the FLH data outputs produced on those locations. The analyst visited each of the five locations two or three times, for a total of twelve experiments. Despite his testing, the analyst did not know the algorithm used in creating FLH data and did not know how various factors were weighed to create FLH data outputs. The analyst also could not explain how the uncertainty radius for a frequent location was determined. He was able to identify that the uncertainty radius and center coordinate point for a frequent location could change with each visit to that location, but he was unable to explain why the uncertainty radius for a frequent location changed or whether data from previous visits contributed to how FLH data changed after a subsequent visit to the location. Moreover, while the FLH data for some locations included a "confidence level," the analyst could not explain what the confidence level meant, why some locations had a confidence level and others did not, or how the confidence level was calculated.

Police recovered the phone eight days after the murder when Arrington himself was shot with another man - who died - in a car on Richfield Street in Dorchester.

Police, however, did not charge Arrington with Long's murder until 2016.

Since then, his trial date has been pushed back by appeals. Prosecutors appealed a Suffolk Superior Court judge's ruling to toss the phone and any data from it as evidence, and in 2020 won, but two days before he was set to go on trial in September of last year, appealed a separate ruling barring their use of the FLH data, leading to today's SJC ruling.