New research has found that some streaming devices and dozens of Android and iOS apps are secretly being used for fraud and other cybercrime.

When you buy a TV streaming box, there are certain things you wouldn’t expect it to do. It shouldn’t secretly be laced with malware or start communicating with servers in China when it’s powered up. It definitely should not be acting as a node in an organized crime scheme making millions of dollars through fraud. However, that’s been the reality for thousands of unknowing people who own cheap Android TV devices.

In January, security researcher Daniel Milisic discovered that a cheap Android TV streaming box called the T95 was infected with malware right out of the box, with multiple other researchers confirming the findings. But it was just the tip of the iceberg. Today, cybersecurity firm Human Security is revealing new details about the scope of the infected devices and the hidden, interconnected web of fraud schemes linked to the streaming boxes.

Human Security researchers found seven Android TV boxes and one tablet with the backdoors installed, and they’ve seen signs of 200 different models of Android devices that may be impacted, according to a report shared exclusively with WIRED. The devices are in homes, businesses, and schools across the US. Meanwhile, Human Security says it has also taken down advertising fraud linked to the scheme, which likely helped pay for the operation.

“They’re like a Swiss Army knife of doing bad things on the internet,” says Gavin Reid, the CISO at Human Security who leads the company’s Satori Threat Intelligence and Research team. “This is a truly distributed way of doing fraud.” Reid says the company has shared details of facilities where the devices may have been manufactured with law enforcement agencies.

Human Security’s research is divided into two areas: Badbox, which involves the compromised Android devices and the ways they are involved in fraud and cybercrime. And the second, dubbed Peachpit, is a related ad fraud operation involving at least 39 Android and iOS apps. Google says it has removed the apps following Human Security’s research, while Apple says it has found issues in several of the apps reported to it.

First, Badbox. Cheap Android streaming boxes, usually costing less than $50, are sold online and in brick-and-mortar shops. These set-top boxes often are unbranded or sold under different names, partly obscuring their source. In the second half of 2022, Human Security says in its report, its researchers spotted an Android app that appeared to be linked to inauthentic traffic and connected to the domain flyermobi.com. When Milisic posted his initial findings about the T95 Android box in January, the research also pointed to the flyermobi domain. The team at Human purchased the box and multiple others, and started diving in.

In total the researchers confirmed eight devices with backdoors installed—seven TV boxes, the T95, T95Z, T95MAX, X88, Q9, X12PLUS, and MXQ Pro 5G, and a tablet J5-W. (Some of these have also been identified by other security researchers looking into the issue in recent months). The company’s report, which has data scientist Marion Habiby as its lead author, says Human Security spotted at least 74,000 Android devices showing signs of a Badbox infection around the world—including some in schools across the US.

The TV devices are built in China. Somewhere before they reach the hands of resellers—researchers don’t exactly know where—a firmware backdoor is added to them. This backdoor, which is based on the Triada malware first spotted by security firm Kaspersky in 2016, modifies one element of the Android operating system, allowing itself to access apps installed on the devices. Then it phones home. “Unbeknownst to the user, when you plug this thing in, it goes to a command and control (C2) in China and downloads an instruction set and starts doing a bunch of bad stuff,” Reid says.

Human Security tracked multiple types of fraud linked to the compromised devices. This includes advertising fraud; residential proxy services, where the group behind the scheme sell access to your home network; the creation of fake Gmail and WhatsApp accounts using the connections; and remote code installation. Those behind the scheme were selling access to residential networks commercially, the company’s report says, claiming to have access to more than 10 million home IP addresses and 7 million mobile IP addresses.

The findings tally with those of other researchers and ongoing investigations. Fyodor Yarochkin, a senior threat researcher at security firm Trend Micro, says the company has seen two Chinese threat groups that have used backdoored Android devices—one it has researched deeply, the other is the one Human Security looked at. “The infection of devices is quite similar,” Yarochkin says.

Trend Micro has found a “front end company” for the group it investigated in China, Yarochkin says. “They were claiming that they have over 20 million devices infected worldwide, with up to 2 million devices being online at any point of time,” he says. Based on Trend Micro’s network data, Yarochkin believes these figures to be credible. “There was a tablet in one of the museums somewhere in Europe,” Yarochkin says, adding he believes it is possible that swaths of Android systems may have been impacted, including in cars. “It’s easy for them to infiltrate the supply chain,” he says. “And for manufacturers, it's really difficult to detect.”

Then there’s what Human Security calls Peachpit. This is an app-based fraud element, which has been present on both the TV boxes as well as Android phones and iPhones, Reid says. The company identified 39 Android, iOS, and TV box apps that were involved. “These are template-based applications—not very high quality,” says Joao Santos, a security researcher at the company. Apps about developing six-pack abs and logging the amount of water a person drinks were included.

The apps performed a range of fraudulent behavior, including hidden advertisements, spoofed web traffic, and malvertising. The research says that while those behind Peachpit appear different from those behind Badbox, it is likely they are working together in some way. “They have this SDK that did the ad fraud part, and we found a version of this SDK that matches the name of the module that was being dropped on the Badbox,” Santos says, referring to a software development kit. “That was another level of connection that we found.”

Human Security’s research says the ads involved were making 4 billion ad requests per day, with 121,000 Android devices impacted and 159,000 iOS devices impacted. There had been 15 million downloads in total for the Android apps, the researchers calculate. (The Badbox backdoor was found only on Android, not on any iOS devices.) Reid says that based on the data the company has, which isn’t a complete picture due to the complexity of the ad industry, those behind the scheme could have easily earned $2 million in one month alone.

Google spokesperson Ed Fernandez confirms the 20 Android apps reported by Human Security have been removed from the Play Store. “The off-brand devices discovered to be Badbox-infected were not Play Protect–certified Android devices,” Fernandez says, referring to Google’s security testing system for Android devices. “If a device isn't Play Protect certified, Google doesn’t have a record of security and compatibility test results.” The company has a list of certified Android TV partners. Apple spokesperson Archelle Thelemaque says that it found five of the apps Human reported breaching its guidelines, and the developers were given 14 days to make them follow the rules. Four of them have done so, as of publication.

Toward the end of 2022 and in the first part of this year, Reid says, Human Security took action against the advertising fraud elements of Badbox and Peachpit. According to data shared by the company, the amount of fraudulent ad requests from the schemes taking place now has completely dropped off. But the attackers adapted to the disruption in real time. Santos says when the countermeasures were first deployed, those behind the schemes started by sending out an update to obfuscate what they were doing. Then, he says, those behind Badbox took down the C2 servers powering the firmware backdoor.

While the attackers have been slowed, the boxes are still in people’s homes and on their networks. And unless someone has technical skills, the malware is very hard to remove. “You can think of these Badboxes as kind of like sleeper cells. They're just sitting there waiting for instruction sets,” Reid says. Ultimately, for people buying TV streaming boxes, the advice is to buy branded devices, where the manufacturer is clear and trusted. As Reid says, “Friends don't let friends plug in weird IoT devices into their home networks.”

Your Cheap Android TV Streaming Box May Have a Dangerous Backdoor

Elon Musk’s brain-chip startup conducted years of tests at UC Davis, a public university. A WIRED investigation reveals how Neuralink and the university keep the grisly images of test subjects hidden.

The tan macaque with the hairless pink face could do little more than sit and shiver as her brain began to swell. The California National Primate Center staff observing her via livestream knew the signs. Whatever had been done had left her with a “severe neurological defect,” and it was time to put the monkey to sleep. But the client protested; the Neuralink scientist whose experiment left the 7-year-old monkey’s brain mutilated wanted to wait another day. And so they did.

As the attending staff sat back and observed, the monkey seized and vomited. Her pupils reacted less and less to the light. Her right leg went limp, and she could no longer support the weight of her 15-pound body without gripping the bars of her cage. One attendant moved a heat lamp beside her to try to stop her shaking. Sometimes she would wake and scratch at her throat, retching and gasping for air, before collapsing again, exhausted.

An autopsy would later reveal that the mounting pressure inside her skull had deformed and ruptured her brain. A toxic adhesive around the Neuralink implant bolted to her skull had leaked internally. The resulting inflammation had caused painful pressure on a part of the brain producing cerebrospinal fluid, the slick, translucent substance in which the brain sits normally buoyant. The hind quarter of her brain visibly poked out of the base of her skull.

Elon Musk says no primates died as a result of Neuralink’s implants. A WIRED investigation now reveals the grisly specifics of their deaths as US authorities have been asked to investigate Musk’s claims.

On September 13, 2018, she was euthanized, records obtained by WIRED show. This episode, regulators later acknowledged, was a violation of the US Animal Welfare Act; a federal law meant to set minimally acceptable standards for the handling, housing, and feeding of research animals. There would be no consequences, however. Between 2016 and 2021, the United States Department of Agriculture (USDA) enforced the humane treatment of animals through what it called “teachable moments.” Because the center—home to a colony of nearly 5,000 primates run by the University of California–Davis—had proactively reported the violation, it could not be legally cited.

And neither could Neuralink. “If you want to split hairs,” a former employee tells WIRED, “the implant itself did not cause death. We sacrificed her to end her suffering.” The employee, who signed a confidentiality agreement, asked not to be identified.

****Got a Tip?****

Missing from the veterinary records released by the university are hundreds of photographs taken by the primate center’s staff between 2018 and 2020 of Neuralink’s test subjects. Though publicly funded, thus bound by California’s open records law, UC Davis has fought disclosure of the photographs for more than a year. Releasing them, it says, would not serve the public’s interest.

Meanwhile, videos of the experiments have seemingly vanished. Documents obtained by WIRED show that the primate center’s staff wrote about reviewing a “tape” of the aforesaid monkey hours before they stopped her heart. The school has not acknowledged that such a tape exists, and Neuralink, whose partnership with the school ended three years ago, was permitted to store its own footage and remove it from the property when it wished.

“They provided their own computing infrastructure, and they had their own network connection, and they have removed their computing infrastructure from the premises,” the school said in a September 2021 email to the Physicians Committee for Responsible Medicine, which is suing UC Davis for the release of images and videos of Neuralink’s experiments there. “The IT staff at the California National Primate Research Center were in no way involved with any aspects of the creation or storage of Neuralink’s video content,” the school added.

Records show that the school ordered Neuralink to request permission before recording any of the animals. And the school reserved the right to view the footage.

Internal emails reviewed by WIRED show that Neuralink, founded and owned by Elon Musk, had tight control over what UC Davis was allowed to divulge about the experiments. Interviews with sources familiar with the tests shed light on the tensions between the school and outside groups over the public’s right to know about research it’s subsidizing.

The sources say secrecy is given top priority, not merely because of the proprietary research being conducted at the center, but also out of fear that the public will respond poorly—perhaps violently—to images of the macaques being experimented on. Though the school’s protocols work effectively to prevent images of the experiments from getting out, it could not legally keep hidden all the written records of Neuralink’s procedures.

Macaques procured for Neuralink from UC Davis’ colony were trained months and even years before going under the knife, a former Neuralink employee recently told WIRED. But the prospect for survival was abysmal for some, they say, due in part to “poor planning and poor procedure.” Early on, the Neuralink researcher says, the company lacked personnel crucial to the operation. “We didn’t have any surgical techs. We didn’t even have a veterinary pathologist on staff at the time.”

After an animal was “sacrificed,” few if any records were created. The former employee claims Neuralink worked purposely to keep records of its work out of UC Davis’ hands—specifically to shield them from public records requests. The products under development at the center are proprietary and created for profit, the researcher says, not to “further the knowledge of mankind.”

Emails obtained by WIRED through a public records request show Davis’s staff scrambling in February 2018—the earliest days of the partnership—to get Neuralink’s equipment up and running. The university had agreed to provide the firm with a dedicated on-site network with a secure uplink to a remote facility. In one email, a faculty member noted that Neuralink had been warned against “live streaming” or producing any “recording of actual monkeys.” Asked if the same rules would apply after Neuralink’s equipment was set up, another Davis official said once installed “they can do whatever they want.”

Neuralink did not respond to WIRED’s request to comment. UC Davis spokesperson Andy Fell maintains that the university has complied with the California Public Records Act, having supplied the “vast majority of records” requested by the Physicians Committee. “Some requested items were not provided because they are exempt from disclosure under the law for various reasons set out in court filings,” he says. “All animal research at UC Davis, including contract research like that performed by Neuralink, is conducted under the same rules and regulations and overseen by the UC Davis Institutional Animal Care and Use Committee (IACUC),” Fell adds.

UC Davis’ Institutional Animal Care and Use Committee did not respond to a request for comment.

Davis has released hundreds of pages of emails, contractual documents, memos, and other veterinary records detailing the public university’s work for Neuralink between 2018 and 2020. The descriptions of botched surgeries and the suffering of the subjects was enough to provoke media investigations and coax comments of concern from a handful of lawmakers.

Hundreds of files remain under lock and key—including photographs of the neurological damage that resulted from Neuralink’s work with the macaques. The experiments involved drilling a hole roughly the size of a US dime into the monkey’s skulls, placing electrodes inside their brains, and screwing titanium plates to their skulls. UC Davis says the value of the photos of these operations now lies exclusively in “informing future research and clinical practices,” or what it calls “the refinement of surgical techniques.”

In October 2022, the Physicians Committee sued UC Davis—a public institution, funded in part by US taxpayers—in an attempt to gain access to records of Neuralink’s work. The Physicians Committee, which aims to promote alternatives to animal testing, has many detractors in the scientific community. The American Medical Association, which supports the use of animals in biomedical research, is one of the largest.

The Physicians Committee has argued in California state court that the public has the right to know about any suffering resulting from taxpayer-funded animal tests. “Disclosure of the footage is particularly important because Neuralink actively misleads the public about, and downplays the gruesome nature of, the experiments,” Corey Page, an attorney with Evans and Page who is representing the Physicians Committee in the lawsuit, tells WIRED.

The Physicians Committee’s suit against UC Davis, filed in California state court in Yolo County, is ongoing.

As it is a public records law that UC Davis is fighting, its arguments against greater transparency are centered around what’s best for the public. According to the school's attorneys, that means the public should not see images of Neuralink’s work.

One researcher familiar with the photos conceded they are particularly gruesome. “A macaque skull with the flesh torn out of it is not a pretty image,” they say. The school routinely deals with protesters, the source says. As a result, any visual evidence of experiments or animal subjects are tightly controlled. Filming the monkeys without the permission of the facility’s director is forbidden. Davis exercises the right to “pre-review” any media it allows to be captured.

A typical request for a recording at the colony, approved in August 2019, aimed to capture how a monkey’s breathing caused “vibration and movement” in a brain implant. Neuralink’s researchers emphasized in paperwork obtained by WIRED that the subject would “NOT be in focus” herself.

Court records show Davis’s attorneys have argued that the most likely outcome of releasing the photos is that its own pathologists will simply stop taking them. While losing a “useful note‐taking and memory‐jogging” tool, they say, refusing to take photos at the necropsy stage of the experiments may also run afoul of federal regulatory guidelines, enforced by the USDA and the campus’s own “animal use” committee. Compliance with this committee, incidentally, is a prerequisite of the center’s federal funding.

A document known as a Vaughn index relays the school’s specific rationale for withholding more than 370 photos, which may be subject to release under the 1968 California Public Records Act. The index lays out the theory that viewing the images would have such a visceral impact on the public that fears of reprisal among Davis’ researchers would be detrimental to their work ethic.

It centers on the belief that the public is too naive at large to distinguish between scientific inquiry and senseless butchery. In its own words, there is a high risk of “non‐contextual misinterpretation of the photos by persons who are not privy to the contextual facts.”

“The interest in protecting the safety of public employees and ensuring research that benefits the public can proceed without risk of violence clearly outweighs the public's interest in viewing said photographs,” the document says.

The risk of public officials being harassed is a factor addressed by most, if not all, open records laws in the United States, which are traditionally built on a presumption that favors disclosure. In most cases, being a mid- to low-level employee is enough to warrant redaction by default–a rule that is generally observed, as well, by professional news organizations. In veterinary records reviewed by WIRED, Davis has consistently censored the names of all of its staff, including those at director level. The university has even redacted identifying information about the animals, including their names and other identifiable information.

In part, Davis argues that the photos represent what’s called a “deliberative” work product. Public officials are often protected from disclosing information that reflects meditations on policies that have yet to become rule or law. The object is to encourage debate and consideration of all ideas—not just the ones assumed to be good from the start. This privilege, however, does not extend to documents related to policies actually enacted. And the school states clearly that the photos it’s choosing to withhold have been kept only to inform future policy.

History shows there are other consequences to the release of photographs specifically with regards to animal treatment. The creation of the US Animal Welfare Act can be tied to the public’s reaction to evidence of animal abuse published more than 60 years ago in _Life_ magazine. Without being able to see what _Life_ audaciously called “concentration camps for dogs,” it’s difficult to predict just how long it would have been before the United States finally adopted a law to protect animals that are bought, sold, and experimented upon.

Neuralink ended its partnership with UC Davis in September 2020, but the Physicians Committee claims that it continues to employ the same neurosurgeon and many of the staff responsible for the experiments that poisoned, maimed, and ultimately killed at least 12 macaque monkeys.

Last month, the company announced that it is preparing to start human trials after receiving a green light from the US Food and Drug Administration. Since ending its partnership with Davis, Neuralink has brought its testing in-house—far from the prying eyes of journalists, animal welfare groups, and the jurisdiction whose records laws first shed light on its practices.

How Neuralink Keeps Dead Monkey Photos Secret

Victims of the MOVEit breach continue to come forward. But the full scale of the attack is still unknown.

In a field of shocking, opportunistic espionage campaigns and high-profile digital attacks on popular businesses, the biggest hack of 2023 isn’t a single incident, but a juggernaut of related attacks that keeps adding victims to its score. In the coming months, more people, as many as tens of millions, could find out that their sensitive information has been compromised. But more still will likely never learn of the situation or its impact on them.

Since May, mass exploitation of a vulnerability in the widely-used file transfer software MOVEit has allowed cybercriminals to steal data from a dizzying array of businesses and governments, including Shell, British Airways, and the United States Department of Energy. Progress Software, which owns MOVEit, patched the flaw at the end of May, and broad adoption of the fix ultimately halted the rampage. But the “Clop” data extortion gang had already orchestrated a far-reaching smash and grab. And months later, the full extent of the damage is still coming into view.

Last week, Ontario’s government birth registry, BORN Ontario, said that it suffered a MOVEit-related attack earlier this year in which hackers stole sensitive personal data from 3.4 million people, including 2 million babies as well as expectant parents and those seeking fertility care. The compromised health data dates from January 2010 to May 2023. While organizations like BORN continue to disclose a slow trickle of MOVEit incidents, researchers say that the number of suspected attacks—and the total number of people whose data has already been stolen in these incidents—far exceeds what has come to light.

“I don’t think we’re done hearing about this by any means. We’re going to keep seeing that rolling disclosure over probably the next few months,” says Emily Austin, security research manager and senior researcher at the threat intelligence firm Censys. “These companies are completing their investigations—they’re starting to notify customers who might have been affected.”

Austin points out that one of the nuances of the MOVEit situation is that it is a true software supply chain security issue. The vulnerabilities existed in two versions of the MOVEit service: the cloud service known as MOVEit Cloud, and the local version that institutions run themselves on their premises, known as MOVEit Transfer. The latter is where most of the exploitation occurred. But many organizations that had data stolen in MOVEit exploitation attacks weren’t directly using it. Instead, they’d collaborated with a third party or contracted with a vendor that does. Attackers were able to steal whatever data they could grab from vulnerable MOVEit systems, whether the information was from one institution or many.

“An advanced and persistent threat actor used a sophisticated, multi-stage attack to exploit this zero-day vulnerability, and we are committed to playing a collaborative role in the industry-wide effort to combat cybercriminals intent on maliciously exploiting vulnerabilities in widely used software products,” Progress Software said in a statement.

Centralized data repositories like MOVEit have been particularly appealing targets to Clop, which is known for strategically exploiting systems embedded in the software supply chain, including multiple file transfer tools. Earlier this year, Clop claimed it breached more than 100 organizations by abusing the GoAnywhere file transfer tool. The gang also mounted a massive data extortion campaign at the end of 2020 by exploiting flaws in Accellion networking equipment.

The MOVEit incident eclipses them, though, both in the number of victim organizations and individuals whose data was compromised. Antivirus company Emsisoft has been tracking the number of MOVEit victim organizations that have publicly declared they were impacted since May. The researchers have combed individual US state breach notifications, filings with the US Securities and Exchange Commission, public disclosures, and Clop's own disclosure website to tabulate and reconcile the true toll of the attacks.

To date, Emsisoft has concluded that 2,167 organizations have been impacted by the sprawling campaign. The number had been hovering around 1,000 in recent months, but it jumped significantly when the National Student Clearinghouse revealed 890 colleges and universities across the US—including Harvard University and Stanford University—had been impacted by MOVEit breaches. Organizations in the US account for 88.8 percent of known victims, according to Emsisoft, while a smattering of other organizations in Germany, Canada, and the UK have also been exposed by Clop and come forward.

According to Emsisoft’s analysis, around 1,841 organizations have disclosed breaches, but only 189 of them have specified how many individuals were impacted by the incident. From these detailed disclosures, Emsisoft has found that more than 62 million individuals had their data breached as part of Clop’s MOVEit spree. But since there are estimated to be nearly 2,000 organizations that have not revealed how many individuals had personal data affected in their breaches—and since researchers have concluded that there are other impacted organizations that haven’t come forward at all—the true total of people whose data was compromised is likely even larger, possibly on the scale of hundreds of millions of individuals, according to Emsisoft.

“It’s inevitable that there are corporate victims that don’t yet know they’re victims and there are individuals out there who don’t yet know they’ve been impacted,” says Brett Callow, a threat analyst at Emsisoft. “MOVEit is especially significant simply because of the number of victims, who those victims are, the sensitivity of the data that was obtained, and the multitude of ways that data can be used.”

Censys’ Austin says file transfer tools are by their nature a “fantastic target” for cybercriminals. The whole purpose of the tools is to manage and share data, so these services are often trusted with large volumes of sensitive information. BORN Ontario said in a statement last week that the data taken in the breach was from those “seeking pregnancy care and newborns.” This included lab test results, pregnancy risk factors, and procedures. Names, dates of birth, government ID numbers like Social Security numbers, addresses, and more have all been compromised in other MOVEit incidents.

While cybercriminal groups often make headlines for attention-grabbing ransomware or extortion attacks, such as those against casinos, persistent and unrelenting theft, publication, extortion, and trade of people’s sensitive data from sprees like the MOVEit rampage can ruin lives—a cumulative reality that is often overshadowed by individual incidents where profits are on the line. Hacks on schools have revealed details of sexual assaults, child abuse allegations, and suicide attempts, with the Associated Press reporting individuals often don’t know the details have been published. Meanwhile, breaches of mental health service providers have exposed patients’ records.

Callows says that he suspects the slow drip of MOVEit-related disclosures “will rumble on for years.” More broadly, he and Austin emphasize that defenders should prepare for cybercriminals to continue targeting widely-used data management software. As Callow puts it, “MOVEIt isn’t the first file transfer application to be exploited and it likely will not be the last.”

Just last week, MOVEit developer Progress Software disclosed a new set of vulnerabilities in one of its file transfer tools for servers, known as WS\_FTP Server, along with patches for the flaws. The company says that it has not “currently” seen evidence that the bugs are being actively exploited.

The Biggest Hack of 2023 Keeps Getting Bigger

A software company sold a New Jersey police department an algorithm that was right less than 1 percent of the time.

_This article was_ _copublished_ _with_ _The Markup__, a nonprofit, investigative newsroom that challenges technology to serve the public good. Sign up for its newsletters_ _here_.

Crime predictions generated for the police department in Plainfield, New Jersey, rarely lined up with reported crimes, an analysis by The Markup has found, adding new context to the debate over the efficacy of crime prediction software.

Geolitica, known as PredPol until a 2021 rebrand, produces software that ingests data from crime incident reports and produces daily predictions on where and when crimes are most likely to occur.

We examined 23,631 predictions generated by Geolitica between February 25 and December 18, 2018, for the Plainfield Police Department (PD). Each prediction we analyzed from the company’s algorithm indicated that one type of crime was likely to occur in a location not patrolled by Plainfield PD. In the end, the success rate was less than half a percent. Fewer than 100 of the predictions lined up with a crime in the predicted category, that was also later reported to police.

Diving deeper, we looked at predictions specifically for robberies or aggravated assaults that were likely to occur in Plainfield and found a similarly low success rate: 0.6 percent. The pattern was even worse when we looked at burglary predictions, which had a success rate of 0.1 percent.

“Why did we get PredPol? I guess we wanted to be more effective when it came to reducing crime. And having a prediction where we should be would help us to do that. I don't know that it did that,” says captain David Guarino of the Plainfield PD. “I don't believe we really used it that often, if at all. That's why we ended up getting rid of it.”

Guarino noted that concerns about accuracy and the department’s general disinterest in using the software suggested that the money paid for Geolitica’s software could have been better spent elsewhere.

The Reading Police Department in Reading, Iowa, introduced Predpol, now rebranded as Geolitica, in 2013.

Photograph: Susan L. Angstadt/Getty Images

“We do have a mentoring program,” Guarino says. “We could probably put \[the money\] toward that, for the summertime. We used to have about 80 kids.” A contract between the department and Geolitica listed a $20,500 subscription fee for its first annual term, and then $15,500 for a yearlong extension.

We sent our findings and methodology to Geolitica prior to publication. Representatives did not respond to multiple requests for comment.

A report from WIRED last week revealed that Geolitica is ceasing operations at the end of this year. SoundThinking, a law enforcement technology company previously known as ShotSpotter, has hired Geolitica’s engineering team, but not its senior leadership, and is in the process of acquiring some of its intellectual property. Geolitica’s current customers are being transitioned into SoundThinking’s patrol platform.

“We believe that offering our law enforcement partners the ability to deploy their limited staffing resources has tremendous merit and that when fully integrated into Resource Router, these highly complementary assets will create a best-in-class deployment system for agencies of all sizes,” SoundThinking senior vice president Sam Klepper wrote in a statement to The Markup.

Information about the shutdown and customer transition to SoundThinking was not posted on Geolitica’s website or social media accounts. SoundThinking also purchased HunchLab, another predictive policing company, from Azavea in 2018.

Klepper explains that SoundThinking plans on incorporating some of Geolitica’s algorithms into its own existing systems, but “the algorithms mentioned do not include any crime prediction modeling source code at all. SoundThinking did not purchase any of Geolitica's crime prediction technology or source code.”

In 2021, The Markup published an investigation in partnership with Gizmodo showing that Geolitica’s software tended to disproportionately target low-income, Black, and Latino neighborhoods in 38 cities across the country. Our investigation was based on data downloaded in January 2021 from an unprotected cloud storage server linked from a page on the Los Angeles Police Department’s website. That server stored predictions Geolitica had provided to police departments in dozens of cities across the country over nearly three years. Access to that server was later restricted.

Our \[The Markup’s\] investigation stopped short of analyzing precisely how effective Geolitica’s software was at predicting crimes because only two out of 38 police departments provided data on when officers patrolled the predicted areas. Geolitica claims that sending officers to a prediction location would dissuade crimes through police presence alone. It would be impossible to accurately determine how effective the program is without knowing which predictions officers responded to and which ones they did not respond to.

After we requested this information from the 38 departments in our original investigation, only Plainfield provided us with more than one day’s worth of data on officer patrol locations.

We tested Geolitica’s accuracy by comparing the system’s predictions to crime reports, which only capture crimes reported to police, not instances when someone was victimized but elected not to inform law enforcement. The Bureau of Justice Statistics reported that less than half of violent and property crimes were reported to police in 2022. Reporting rates are also inconsistent across demographic groups. The agency found in a 2012 report that Black, Latino, and lower-income crime victims were more likely to report the crime than White people and people from higher-income households.

Plainfield police officers reported visiting 129 of the 23,760 prediction locations provided by Geolitica, according to the data. To conduct our analysis, we filtered out those predictions to sidestep the issue of police presence deterring potential crimes, and compared the rest against a list of crime incident reports obtained through public records requests. Each prediction was assigned to one of the department’s four daily patrol shifts, and each shift lasted 11 hours and 15 minutes. If a crime occurred in that location during the shift period, we counted it as a correct prediction.

The lengths of these prediction windows was a major problem, American University law professor Andrew Ferguson, who authored the book _The Rise of Big Data Policing,_ tells The Markup after being informed of our findings. "The timing piece is a flawed system,” he says. “The precision of it was supposed to be having the officers at the right time at the right place, not ‘during an entire day’s shift, there will be crime.’”

Plainfield officials say they never used the system to direct patrols. Instead, officials say all of the instances when the automated monitoring system tagged police vehicles as having visited a prediction location were coincidental overlaps that naturally occurred as officers regularly traveled across the geographically small, largely suburban community. To account for all possibilities, The Markup ran its analysis with the patrolled prediction locations both included and filtered out, and did not see a significant difference in the results.

Records showed that none of the five arrests that occurred in prediction locations during the time period of our analysis could have conceivably been due to officers responding to a Geolitica prediction, which Guarino also confirmed. Similarly, we asked the other 37 departments in our previous investigation if they could point to any arrests that came as a direct result of a Geolitica prediction, but none could. Representatives from neither the National Association of Criminal Defense Lawyers nor the National District Attorneys Association could recall a case going to trial where the arrest came as a direct result of an algorithmic crime prediction.

Despite Plainfield officials’ assertions that the software wasn’t used to inform where officers went, there is a quote on Geolitica’s website from former Plainfield police officer sergeant Larry Brown hailing the program. “Robbery at a restaurant ... PredPol prediction and robbery was 10 ft. away,” wrote Brown. “I train our 100+ officers on how to patrol PredPol boxes and I love our successes.”

In an email to The Markup, Guarino wrote that he wasn’t familiar with the incident Brown, who retired in 2018, mentioned and didn’t “know how Sgt. Brown came to that conclusion.” Multiple attempts to contact Brown were unsuccessful.

A major problem with Geolitica’s system, as it was used in Plainfield, is that there were a massive number of predictions compared to a relatively small number of crimes. In a 2021 email to The Markup regarding our previous investigation into the company’s algorithm, Geolitica CEO Brian MacDonald wrote that the number of predictions generated in each area was set by the law enforcement agency using the software. The data we received for Plainfield, which has a population of around 54,000, had 80 predictions per day for just a handful of crime types, while the maximum number of crimes of any type reported in a single day was 22 during the analysis period. This volume of predictions sits below the median of 107 daily predictions for the 37 other jurisdictions we analyzed in our original investigation. Some larger cities had far more predictions, like Los Angeles, which averaged just over a thousand daily. But there were cities smaller than Plainfield, like Niles, Illinois, which has around half Plainfield’s population, with a large volume of daily predictions—231, on average.

In her 2019 master’s thesis for the Naval Postgraduate School, Ana Lalley, police chief of Elgin, Illinois, wrote critically about her department’s experience with the software, which left officers unimpressed. “Officers routinely question the prediction method,” she wrote. “Many believe that the awareness of crime trends and patterns they have gained through training and experience help them make predictions on their own that are similar to the software’s predictions.”

Lalley added that when the department brought those concerns to Geolitica, the company warned that the software “may not be particularly effective in communities that have little crime.” Elgin, a Chicago suburb, has about double Plainfield’s population.

“I think that what this shows is just how unreliable so many of the tools sold to police departments are,” says Dillon Reisman, founder of the American Civil Liberties Union of New Jersey’s Automated Injustice Project. “We see that all over New Jersey. There are lots of companies that sell unproven, untested tools that promise to solve all of law enforcement’s needs, and, in the end, all they do is worsen the inequalities of policing and for no benefit to public safety.”

David Weisburd, a criminologist who served as a reviewer on a 2011 academic paper coauthored by two of Geolitica’s founders, recalls approving their ideas around crime modeling at the time, but warns that inaccurate predictions can have their own negative externalities outside of wasting officers’ time.

“Predicting crimes in places where they don’t occur is a destructive issue,” Weisburd says. “The police are a service, but they are a service with potential negative consequences. If you send the police somewhere, bad things could happen there.”

One study found that adolescent Black and Latino boys stopped by police subsequently experienced heightened levels of emotional distress, leading to increased delinquent behavior in the future. Another study found higher rates of use of force in New York City neighborhoods led to a decline in the number of calls to the city’s 311 tip line, which can be used for everything from repairing potholes to getting help understanding a property tax bill.

“To me, the entire benefit of this type of analysis is using it as a starting point to engage police commanders and, when possible, community members in larger dialog to help understand exactly what it is about these causal factors that are leading to hot spots forming,” says Northeastern University professor Eric Piza, who has been a critic of predictive policing technology.

For example, the city of Newark, New Jersey, used risk terrain modeling (RTM) to identify locations with the highest likelihood of aggravated assaults. Developed by Rutgers University researchers, RTM matches crime data with information about land use to identify trends that could be triggering crimes. For example, the analysis in Newark showed that many aggravated assaults were occurring in vacant lots.

The RTM then points to potential environmental solutions that come from across local governments, not just police departments. A local housing organization used that New Jersey data to prioritize lots to develop for new affordable housing that could not only increase housing stock but also reduce crime. Other community groups used the crime-risk information to convert city-owned lots to well-lighted, higher-trafficked green spaces less likely to attract crime.

Newark police are seen at the scene after a fatal shooting on August 17, 2023.

Photograph: Kyle Mazza/Getty Images

Even Geolitica seems to be moving away from its original core functionality of crime predictions and toward pitching itself as a broader platform for managing police department data. Both Geolitica’s patrol management platform, and the one operated by SoundThinking, advertise that they can incorporate elements of risk terrain modeling into their systems.

A pitch presentation, prepared for the police department in San Ramon, California, last year and obtained through a public records request, offered three tiers of service the department could purchase, with only the highest-cost option providing “automated daily patrol guidance.” The lower levels offered services designed to help departments better understand their operations, like creating heat maps showing the frequency of officer patrols in each part of a jurisdiction and allowing the department to independently set their own locations for officers to patrol. San Ramon Police Department records manager Jessica Simonds tells The Markup that the department subscribed to the lowest level, without the automated predictions.

A Geolitica patrol heat map showing the frequency of officer patrols in parts of Elgin, Illinois.

Courtesy of San Ramon Police Department

“The problem with predictive policing is the policing part,” says Ferguson. “The data about where crime occurs and where victimization occurs, that's something that should be studied. But that doesn't mean that the solution is policing. It might mean the solution is figuring out why people are victimized and funding the solution to that.”

Predictive Policing Software Terrible at Predicting Crimes

Checking out this AI chatbot's new features? Make sure to keep these privacy tips in mind during your interactions.

With its most recent update, Google Bard can now sort through your trove of Google Docs, rediscover ancient Gmail messages, and search through every video on YouTube. Before experimenting too much with the new extensions available for Google’s chatbot, it's worth going over the steps you can take to protect your privacy (and the ones you can't).

Google Bard launched in March of this year, one month after OpenAI released ChatGPT to the public. You’re likely familiar with how chatbots are designed to mimic human conversation, but Google’s latest features are designed to give Bard more practical applications and uses.

But when every conversation you have with Bard is tracked, logged, and used again to train the AI, how can you trust it with your data? Here are some tips to protect your prompts and assert some control over what information you give Bard. Then we’ll discuss location data, for which Google sadly provides fewer privacy options.

The default Bard setting is to store every interaction you have with the chatbot for 18 months. In addition to your prompts, Bard stores your approximate location, IP address, and any physical addresses connected to your Google account for work or home. While the default settings are activated, any conversation you have with Bard could be selected for human review.

Want to turn this off? When you’re in the Bard Activity tab, you can stop it from autosaving your prompts and also delete any past interactions. “We give you this choice around Bard Activity, which you can turn on or off, if you want to keep your conversations \[from being\] human-reviewable,” says Jack Krawczyk, a product lead at Google for Bard.

After you turn off Bard Activity, your new chats are not submitted for human review, unless you report a specific interaction to Google. But there’s a catch: If you turn off Bard Activity, then you can’t use any of the chatbot's extensions connecting the AI tool to Gmail, YouTube, and Google Docs.

Yes, you can choose to delete interactions with Bard manually, but the data might not be removed from Google servers until some point later, when the company decides to delete it (if ever). “To help Bard improve while protecting your privacy, we select a subset of conversations and use automated tools to help remove personally identifiable information,” a Google support page reads. The conversations selected for human review are no longer tied to your personal account, and these interactions are stored by Google for up to three years, even if you clear it out of your Bard Activity.

It's also worth noting that any Bard conversation you want to share with friends or coworkers could potentially be indexed by Google Search. At the time of publication, multiple Bard interactions were accessible through Search, from a job hunter requesting advice when applying for a position at YouTube Music to someone asking for 50 different ingredients they could blend into protein powder.

To remove any Bard links you’ve shared, go to **Settings** in the top right corner, select **Your public links**, and click the trash icon to stop online sharing. Google posted on social media that it's taking steps to stop shared chats from being indexed by Search.

This may make you wonder: If I’m using Bard to find my ancient emails, do those conversations remain private? Maybe, maybe not. “With the ability to have Bard summarize and extract content from your Gmail and your Google Docs, we took it a step even further,” says Krawczyk. “Nothing from there is ever eligible. No matter what settings you have turned on. Your email will never be read by another human. Your Google Docs will never be read by another human.” Although no human readers may sound like a bit of relief, it remains unclear how Google uses your data and interactions to train their algorithm or future iterations of the chatbot.

OK, now what about your location data? Are there any tools to limit when Bard keeps track of where you are? In a pop-up, Bard users are given the option whether or not to share their precise location with the chatbot. Even if you opt out of precise location sharing, Bard will still know where you’re at.

“Location data is always collected if you use Bard so that Bard can provide you with a response that is relevant to your query,” reads a Google support page. The chatbot uses your IP address, a tag that reveals your general location, and any personal addresses saved to your Google account. The company claims to anonymize the location data you provide by storing it with at least 1,000 other people’s information in a tracking area that’s 2 miles or larger.

While location tracking may make you uncomfortable, companies keep track of IP addresses to know their users’ locations more commonly than you might realize. For example, Google Search uses your IP address (and other sources) to help answer all those “near me” queries: best takeout near me, used camping gear near me, movie theater tickets near me. Of course, just because it’s common doesn’t make it okay for everyone. Keep that in mind when you use products like Bard.

Even though Google does not provide an easy way for you to opt out of Bard’s location tracking, one method to mask your IP address is to use a virtual private network. These tools are available for your PC as well as your mobile device. To find the best VPN for your particular situation, check out WIRED’s roundup of the best options.

How to Stop Google Bard From Storing Your Data and Location

Every smartphone has an expiration date. Here’s when yours will probably come.

If you're shopping for a smartphone, you're probably weighing how powerful it is, how good the cameras are, and of course how much you're going to have to pay for it—but it's also worth considering how long the handset is going to last you.

A big part of that calculation comes down to the length of time that the phone will get updates. Apple just pushed out iOS 17, a software update that is heading to iPhones including the iPhone XR and the iPhone XS—handsets that launched in 2018.

For five-year-old phones to be getting (mostly) the same software features as the brand new iPhone 15 is something Apple can be proud of and that its users can be grateful for, but this kind of future-proofing isn't standard.

At the time of writing, Google promises Android updates for its Pixel phones for at least three years. For flagship Samsung Galaxy phones, the software update guarantee is for four years, and for the latest Fairphone 5, it's five years.

Pixel phones are covered for three years of software updates.

Courtesy of Google

There's another important time span to consider though, which is the one for security updates. These updates continue to be issued for older phones even after major software updates have finished, so important vulnerabilities can be patched and phones can be kept safe even if they're no longer getting new and improved features.

For Google and Samsung, security updates are provided for five years, while Fairphone is promising at least eight for the Fairphone 5, and possibly as many as 10. Apple doesn't have a fixed approach but tends to issue security updates for a year or two after software updates have finished—after seven years, Apple products are declared obsolete.

You may be planning to upgrade your smartphone before that happens, but not everyone will be—and knowing how long your device is going to be supported with updates from its manufacturer then becomes very important.

If you're struggling to find out this information about a particular gadget, the web can help—and specifically the endoflife.date site, which lists a whole host of hardware devices and software packages. You can see continually updated countdowns for manufacturer support for iPhones, Pixels, Galaxy phones, and more.

When Security Updates End

Software updates keep your phone properly patched up and supplied with the latest new features. In contrast, security updates make sure that a device stays secure and protected against evolving threats even after software updates have stopped. Which then prompts the question: What happens when security updates end?

To begin with, your device isn't going to suddenly stop working and become a useless brick of electronics. Assuming the components inside it are still functioning—which is by no means a given after five or six years—then you can continue to use the device in the same way that you always have.

As time goes on, you might start to notice a problem with some of your apps: Apps will often need the very latest versions of iOS or Android in order to run, and if you're not getting those updates, they may stop working or start misbehaving, or you may find yourself unable to update to the latest versions.

Security is a bigger problem. There will be some legacy protection in terms of the patches you've applied over the lifetime of the device, but your phone is no longer going to be prepared for newer exploits and malware attacks, and that puts your data at risk.

iOS 17 works on the iPhone 15 Pro—and on iPhones from 2018.

Courtesy of Apple

When mobile security vulnerabilities emerge, you'll find that many of them are aimed at older versions of operating systems, versions that don't have all of the necessary defense measures in place. The more time that goes by since the last security update, the bigger the risk you're running by still using your phone.

Apple, Google, and Samsung will often scramble to fix security issues as they're discovered, pushing out updates outside of the regular schedule in order to keep devices protected—and again, your phone is ineligible for these patches, except under rare circumstances, if you've reached the end-of-life stage.

You can choose to keep on using your phone: The usual practices of being careful with apps and websites, turning on two-factor authentication for your accounts, and keeping the lock screen securely locked all still apply, and offer some level of protection. However, your device is at significantly more risk of an attack.

As expensive as it may be, it might be time to think about upgrading your phone so you keep receiving updates—it doesn't have to be a brand-new one to get the latest iOS or Android software. When you do decide to say goodbye to your old handset, be sure to dispose of it responsibly.

How to Tell When Your Phone Will Stop Getting Security Updates

Plus: Stolen US State Department emails, $20 million zero-day flaws, and controversy over the EU’s message-scanning law.

WIRED broke the news on Wednesday that SoundThinking, the company behind the gunshot-detection system ShotSpotter, is acquiring some assets—including patents, customers, and employees—from the firm Geolitica, which developed the notorious predictive policing software PredPol. WIRED also exclusively reported this week that the nonprofit Electronic Privacy Information Center is calling on the US Justice Department to investigate potentially biased deployment of ShotSpotter in predominantly Black neighborhoods.

As the US federal government inches closer to a possible shutdown, we took a look at the sprawling conservative media apparatus and deep bench of right-wing hardliners in Congress that are exploiting their leverage to block a compromise in the House of Representatives.

Satellite imaging from the Conflict Observatory at Yale University is providing harrowing insight and crucial information about the devastation wrought in the city of Khartoum by Sudan’s civil war. Meanwhile, researchers from the cybersecurity firm eQualitie have developed a technique for hiding digital content in satellite TV signals—a method that could be used to circumvent censorship and internet shutdowns around the world. And the productivity data that corporations have increasingly been gathering about their employees using monitoring software could be mined in an additional way to train AI models and eventually automate entire jobs.

Plus, there's more. Each week, we round up the security and privacy news we didn’t cover in depth ourselves. Click the headlines to read the full stories, and stay safe out there.

A China-linked hacking group, dubbed BlackTech, is compromising routers in the US and Japan, secretly modifying their firmware and moving around company networks, according to a warning issued by cybersecurity officials this week. The United States Cybersecurity and Infrastructure Security Agency (CISA), the NSA, FBI, and Japan's National Police Agency and cybersecurity office issued the joint alert saying the BlackTech group was “hiding in router firmware.”

The officials said they had seen the Chinese-linked actors using their access to the routers to move from “global subsidiary companies” to the networks of companies’ headquarters in the US and Japan. BlackTech, which has been operating since around 2010, has targeted multiple router types, the officials said, but they highlighted that it compromised Cisco routers using a customized backdoor. “TTPs against routers enable the actors to conceal configuration changes, hide commands, and disable logging while BlackTech actors conduct operations,” the alert says.

Microsoft and US government officials said in July that Chinese government hackers had breached the cloud-based Outlook email systems of about 25 organizations, including the US State Department and Department of Commerce. On Wednesday, an anonymous staffer for Senator Eric Schmitt told Reuters that the State Department incident exposed 60,000 emails from 10 accounts. Nine of the accounts were used by State Department employees focused on East Asia and the Pacific, while one was focused on Europe. The Congressional staffer learned the information in a State Department IT briefing for legislators and shared the details with Reuters via email.

The zero-day market, where new vulnerabilities and the code needed to exploit them are traded for cash, is big business. And it is, maybe, getting more lucrative. Russian zero-day seller Operation Zero this week announced it would increase some of its payments from $200,000 to $20 million. “As always, the end user is a non-NATO country,” the group said, indicating it means Russian private and government organizations.

Unlike bug bounties, where security researchers find flaws in companies’ code and then disclose them to the firms to fix for payments, the zero-day market encourages the trade in flaws that can potentially be exploited by the purchasers. “Full chain exploits for mobile phones are the most expensive products right now and they’re used mostly by government actors," Operation Zero CEO Sergey Zelenyuk told TechCrunch. "When an actor needs a product, sometimes they’re ready to pay as much as possible to possess it before it gets into the hands of other parties.”

The European Union's proposed law to clamp down on child sexual abuse content—by scanning people’s messages and potentially compromising encryption—is one of the continent's most controversial laws of the last decade. This week, a series of revelations from a group of reporters has shown how the law’s main architect was heavily lobbied ahead of proposing the law and that police wanted access to the message data. First, an investigation revealed the close connections between the European Union’s home affairs commissioner, Ylva Johansson, and child protection groups. A second report shows the European police agency Europol pushed to get access to data collected under the proposed law. In response to the investigations, Europe's Committee on Civil Liberties, Justice, and Home Affairs has written to Johansson asking questions about the relationships.

Chinese Hackers Are Hiding in Routers in the US and Japan

Plus: Mozilla patches 10 Firefox bugs, Cisco fixes a vulnerability with a rare maximum severity score, and SAP releases updates to stamp out three highly critical flaws.

Fall is here, but hot zero-day summer shows no signs of cooling down, with the likes of Apple, Microsoft, and Google fixing flaws being used in real-life attacks.

Some major enterprise fixes were released during the month, including a Cisco patch for a vulnerability with a maximum CVSS score of 10.

Spyware has been a prominent trend over the past couple of months, and it’s a threat everyone should take seriously. Attacks can reach devices without any interaction from the user, so it’s important to keep your operating system up to date.

Here’s everything you need to know about the patches issued in September.

Apple iOS and iPad OS

Apple didn’t release any security updates in August, but the iPhone maker sure made up for it in September. First came iOS 16.6.1, an emergency security update released on September 9 to fix two flaws already being used in so-called “zero-click” attacks.

Reported by researchers at the University of Toronto’s Citizen Lab, the vulnerabilities were used to plant spyware via attachments containing malicious images in an iMessage, in an attack the researchers called BLASTPASS.

In mid-September, Apple released its major software upgrade, iOS 17, followed by iOS 17.0.1 a few days later. The surprise iOS 17.0.1 upgrade was important because it fixed another three iPhone flaws being used in spyware attacks.

Tracked as CVE-2023-41992 and reported by security researchers at Citizen Lab and Google, the first issue is a flaw in the kernel that could allow an attacker to escalate privileges. The other two vulnerabilities in WebKit and Security could potentially be chained together to take over a user’s device.

The flaws patched in iOS 17.0.1 were also fixed in the iOS 16.7 release for users of older iPhones or people who don’t want to upgrade to the latest software.

At the end of September, Apple released iOS 17.0.2 to fix some early iOS 17 bugs, and this is the latest version of the software, as of publication.

Apple has also released macOS Sonoma 14 to fix over 60 vulnerabilities.

Google Android

September was a big security month for Google’s Android users, with the monthly patch fixing 33 flaws, including one already being used in attacks. Tracked as CVE-2023-35674, the vulnerability in the framework could allow an adversary to elevate privileges without any interaction from the user. “There are indications that CVE-2023-35674 may be under limited, targeted exploitation,” Google said in its Android Security Bulletin.

Another severe issue is a critical security vulnerability in the system component that could lead to remote code execution with no additional execution privileges needed.

The Android security update has already reached Google’s own Pixel devices as well as Samsung devices, including the Galaxy S23 and S22 series.

Google Chrome

At the end of September, Google updated its Chrome browser after fixing 10 vulnerabilities, one of which was already being exploited by adversaries. Tracked as CVE-2023-5217 and rated as having a high impact, the already-exploited bug is a heap buffer overflow flaw in vp8 encoding in libvpx. “Google is aware that an exploit for CVE-2023-5217 exists in the wild,” the software giant said in an advisory.

The flaw was used in targeted spyware attacks, Google security researcher Maddie Stone later confirmed on Twitter.

Earlier in the month, Google fixed another zero-day flaw, a heap buffer overflow issue initially tracked as CVE-2023-4863, which it thought impacted only the Chrome browser. But two weeks after fixing the issue, researchers discovered it was worse than they thought, affecting the widely-used libwebp image library for rendering images in the WebP format.

Now tracked as CVE-2023-5129, it is thought the bug impacts every application that uses the libwebp library to process WebP images. “The scope of this vulnerability is much wider than initially assumed, affecting millions of different applications worldwide,” security firm Rezilion wrote in a blog.

The security outfit also thinks it is “highly likely” that the underlying issue in the libwebp library is the same issue resulting in CVE-2023-41064—one of the Apple flaws used as part of the BLASTPASS exploit chain to deploy the NSO Group’s Pegasus spyware.

Microsoft

Microsoft’s September Patch Tuesday is one to remember, as it fixed around 65 flaws, two of which are already being exploited by attackers. Tracked as CVE-2023-36761, the first is a Microsoft Word information disclosure vulnerability that could allow NTLM hashes to be exposed.

The second and most severe flaw is a privilege-escalation vulnerability in Microsoft Streaming Service Proxy tracked as CVE-2023-36802. An attacker who successfully exploited this vulnerability could gain system privileges, Microsoft said, adding that exploitation of the flaw has been detected.

Both flaws are marked as important, so it’s a good idea to update your devices as soon as you can.

Mozilla Firefox

Firefox has had a hectic month after Mozilla fixed 10 flaws in its privacy-conscious browser. CVE-2023-5168 is an out-of-bounds write bug in FilterNodeD2D1 affecting Firefox on Windows, rated as having a high impact.

CVE-2023-5170 is a flaw that could result in memory leak from a privileged process. This could be used to effect a sandbox escape if the correct data was leaked, Firefox owner Mozilla said in an advisory.

Meanwhile, CVE-2023-5176 covers memory safety bugs fixed in Firefox 118, Firefox ESR 115.3, and Thunderbird 115.3. “Some of these bugs showed evidence of memory corruption and we presume that with enough effort some of these could have been exploited to run arbitrary code,” Mozilla said.

Cisco

At the start of the month, Cisco issued a patch for a vulnerability in the single sign-on implementation of Cisco BroadWorks Application Delivery Platform and Cisco BroadWorks Xtended Services Platform that could allow an unauthenticated, remote attacker to forge credentials to access an affected system. Tracked as CVE-2023-20238, the flaw has been given a maximum CVSS score of 10.

Also this month, Cisco patched a zero-day in Adaptive Security Appliance and Firepower Threat Defense software already exploited in Akira ransomware attacks. Tracked as CVE-2023-20269 and with a medium severity CVSS score of 5, the vulnerability in the remote access VPN feature of Cisco Adaptive Security Appliance (ASA) Software and Cisco Firepower Threat Defense (FTD) Software could allow an unauthenticated, remote attacker to conduct a brute-force attack to identify valid username and password combinations.

SAP

Enterprise software firm SAP has issued several important fixes as part of its September Security Patch Day. This includes a patch for CVE-2023-40622, an information disclosure vulnerability in SAP BusinessObjects Business Intelligence Platform with a CVSS score of 9.9. “A successful exploit provides information that can be used in subsequent attacks, leading to a complete compromise of the application,” security firm Onapsis said.

CVE-2023-40309 is a missing authorization check issue in SAP CommonCryptoLib with a CVSS score of 9.8. The flaw can result in an escalation of privileges and in the worst case, attackers can compromise the affected application completely, Onapsis said.

Meanwhile, CVE-2023-42472 is an insufficient file type validation flaw in SAP BusinessObjects Business Intelligence Platform with a CVSS score of 8.7.

Apple, Microsoft, and Google Just Fixed Multiple Zero-Day Flaws

A civil liberties group has asked the DOJ to investigate deployment of the ShotSpotter gunfire-detection system, which research shows is often installed in predominantly Black neighborhoods.

The United States Justice Department (DOJ) is being asked to investigate whether a gunshot-detection system widely in use across the US is being selectively deployed to justify the over-policing of mainly Black neighborhoods, as critics of the technology claim.

Attorneys for the nonprofit Electronic Privacy Information Center—a leading US-based civil liberties group—argue that “substantial evidence” suggests American cities are disproportionately deploying an acoustic tool known as ShotSpotter in majority-minority neighborhoods. Citing past studies, EPIC alleges that data derived from these sensors has encouraged some police departments to spend more and more time patrolling areas where the fewest number of white residents live—an allegation disputed by SoundThinking, the system’s manufacturer.

In a letter today to Merrick Garland, the US attorney general, attorneys for EPIC call for an investigation into whether cities using ShotSpotter are running afoul of the Civil Rights Act—namely, Title VI, which forbids racial discrimination by anyone who receives federal funds.

“State and local police departments around the country have used federal financial assistance to facilitate the purchase of a slew of surveillance and automated decision-making technologies, including ShotSpotter,” EPIC says. Despite mounting evidence of ShotSpotter’s discriminatory impact, there is no indication that its Title VI compliance has ever been seriously assessed. 

A spokesperson for SoundThinking says a statement by the company is forthcoming.

ShotSpotter has been deployed in more than 150 cities in the US, according to the company. It relies on internet-connected acoustic sensors, often attached to utility poles, and aims to detect gunfire using machine algorithms. SoundThinking says “acoustic experts” are on staff around the clock to review alerts and “ensure and confirm that the events are indeed gunfire.” The company claims its sensors have a 97 percent accuracy rate, disputing reports that alerts triggered by fireworks and other high-impact sounds affect the system’s accuracy.

EPIC is urging the DOJ to consider research that suggests ShotSpotter has produced “tens of thousands of false alerts” while simultaneously being deployed “in predominantly Black neighborhoods.” One such study, launched by the city of Chicago’s inspector general in 2021, noted that the “frequency of ShotSpotter alerts in a given area may be substantively changing policing behavior.” The agency concluded that, despite the city’s investment of $23-33 million, ShotSpotter alerts “rarely produce evidence of a gun-related crime, rarely give rise to investigatory stops, and even less frequently lead to the recovery of gun crime-related evidence during an investigatory stop.”

News investigations in Ohio and Texas have similarly raised doubts about the system's effectiveness, revealing that in some cases its alerts have delayed responses to 911 calls. Working to expand the use of ShotSpotter in Houston in late 2020, the city also green-lit a pilot program that saw sensors deployed across two areas where communities are between 80 and 95 percent people of color.

EPIC is pressing Garland to investigate whether local law enforcement agencies have used federal grant money to buy ShotSpotter, and if so, ascertain whether those grants conformed with Title VI. Moreover, EPIC is seeking new guidelines for funding systems designed to automate police work: rules to ensure such arrangements are “transparent, accountable, and nondiscriminatory.” The attorney general should take additional steps, EPIC says, to ensure agencies dispersing federal funds are careful to assess whether tech companies meet “minimum standards of nondiscrimination” and that new police technologies are not only justified but necessary to achieve a “defined goal.”

The attorney general’s office did not immediately respond to a request for comment.

Ron Wyden, a leading privacy hawk in the US Senate, tells WIRED that he planned to urge Garland to adopt EPIC’s recommendations.

“There is more than enough evidence at this point to conclude that technologies like ShotSpotter do essentially nothing to stop crime,” Wyden says, “but instead have a well-documented discriminatory impact on marginalized and vulnerable communities.”

A spokesperson for SoundThinking told WIRED earlier this week that ShotSpotter has previously implemented safeguards meant to improve “impartiality” with regard to patrols, as well as other features designed to curb over-policing. The algorithms that drive patrol recommendations use “objective, noncrime data that mitigates potential bias,” the company said, adding that no personally identifiable information is used to determine where patrols are sent.

WIRED reported yesterday that SoundThinking has been quietly acquiring parts of Geolitica—formerly known as PredPol—the maker of an eponymous “predictive policing” tool. Researchers and critics have accused PredPol of perpetuating racial biases in policing by recycling historic crime data that was built on discriminatory police practices.

The danger, critics say, is that data- and AI-driven policing may prove an effective tool for “laundering” racist police conduct, granting it a veneer of scientific authority.

In 2021, Gizmodo and The Markup coauthored an investigation into PredPol and found that in many cases PredPol’s predictions predominantly targeted Black and Latinx residents. In a majority of the jurisdictions where the tool was being used, PredPol had urged police to mainly patrol neighborhoods in which the cities’ poorest people reside. ( The authors of this story were also members of the team that previously investigated PredPol, and EPIC cites the investigation and another Gizmodo report by an author of this story in its letter.)

“The possibility that ShotSpotter data would be an input for predictive policing tools raises yet another huge red flag,” says Chris Baumohl, an EPIC law fellow. “ShotSpotter has well-documented accuracy issues, and its sensors are placed based on historical crime data, baking in discriminatory policing patterns. Laundering this ‘dirty data’ through predictive policing tools only compounds this discriminatory impact.”

Following the Gizmodo-Markup investigation into PredPol, the Justice Department conceded in 2022 that it was unaware of how often the government was helping local police acquire “predictive policing” technology. For this purpose, it said, the DOJ has no “specific records” of how federal funding is used.

US Justice Department Urged to Investigate Gunshot Detector Purchases

SoundThinking is purchasing parts of Geolitica, the company that created PredPol. Experts say the acquisition marks a new era of companies dictating how police operate.

SoundThinking, the company behind the gunshot-detection system ShotSpotter, is quietly acquiring staff, patents, and customers of the firm that created the notorious predictive policing software PredPol, WIRED has learned.

In an August earnings call, SoundThinking CEO Ralph Clark announced to investors that the company was negotiating an agreement to acquire parts of Geolitica—formerly called PredPol—and transition its customers to SoundThinking’s own “patrol management” solution.

“We have already hired their engineering team,” Clark said during the call, a transcript of which is public. He added that the acquisition of patents and staff would “facilitate our application of AI and machine learning technology to public safety.”

SoundThinking’s absorption of Geolitica marks its latest step in becoming the Google of crime fighting—a one-stop shop for policing tools. Experts who study law enforcement use of technology say the bundling of two controversial technologies signals a new era for the cop-tech industry and has the potential to shape the future of policing in the United States. And while SoundThinking has rebranded “predictive policing” as resource management for police departments, a WIRED analysis of one of the company’s apps found that crime-forecasting technology remains one of its key offerings.

“As a moment of tech history, the purchase is significant,” Andrew Ferguson, an American University law professor and author of _The Rise of Big Data Policing_, tells WIRED. “We are in a consolidation moment with big police tech companies getting bigger, and this move is one step in that process.”

PredPol was one of the first, and perhaps the most widely used, predictive policing algorithms in the United States. Its name, a portmanteau of “predictive policing,” became synonymous with the practice.

The software was developed in 2011 and uses historical crime incident reports to produce daily predictions for where future crime is likely to occur. For years, critics and academics have argued that since the PredPol algorithm relies on historical and unreliable crime data, it reproduces and reinforces biased policing patterns. In December 2021, Gizmodo and The Markup analyzed millions of Geolitica’s crime predictions that were discovered on an unsecured server and found that the software disproportionately—and often relentlessly—targeted low-income communities of color for additional patrols.

In recent years, police departments have dropped PredPol after ultimately finding it ineffective. In 2019, a report by the Los Angeles Police Department’s inspector general found that it was unclear whether PredPol had any effect on crime trends. The LAPD, which was the earliest adopter of PredPol, and had even partnered with researchers to develop the technology, dropped the product in 2020, citing budget costs.

Brian MacDonald, the CEO of Geolitica, declined an interview and did not answer specific questions about the acquisitions. A third-party spokesperson for SoundThinking, Rob Merritt, tells WIRED that Geolitica is ceasing operations at the end of the year.

Consolidation Game

Founded in 1996, SoundThinking is now worth around $232 million. Its flagship product, ShotSpotter, is a gunshot-detection system that uses microphones mounted on traffic signals and light poles to detect and locate possible gunfire sounds. For years, activists and academics around the US have fought against the expansion of ShotSpotter, claiming that it is not only inaccurate but is deployed disproportionately in non-white neighborhoods.

Two investigations by local media—_The Houston Chronicle_ and Southwest Ohio’s WYSO—found that ShotSpotter alerts largely resulted in dead ends for police and, in some cases, delayed response times for other calls for service. In 2021, the MacArthur Justice Center at the Northwestern University School of Law analyzed records kept by Chicago’s Office of Emergency Management and Communications over a two-year period and found that 89 percent of ShotSpotter alerts in the city did not lead to police finding evidence of a gun-related crime.

Source

Wired