Gautham Thomas

home

authors

gautham-thomas

In March, Elaine Herzberg walked her bike across the street in Tempe, Arizona, unaware of the Uber self-driving car moving toward her at 38 miles per hour. The car, with a human driver behind the wheel but not in control, struck Herzberg. She later died in a hospital. It was perhaps the first fatality of an unconnected pedestrian involving a self-driving car. Herzberg’s daughter and husband were preparing to file a lawsuit. But 10 days later, Uber settled with Herzberg’s family for an undisclosed sum.

After the accident, Arizona Governor Doug Ducey suspended the testing of Uber self-driving cars in the state, according to a letter he sent to the company. At the time, he said, about 600 vehicles with “automated driving systems” from several companies were cruising Arizona roads. The National Transportation Safety Board (NTSB) is still investigating the incident. A preliminary report was expected to be published in May, but the final examination could take two years to complete, said Christopher O’Neil, a spokesman for the board.

There have been only a handful of fatal accidents involving self-driving cars, so case law guiding liability has not been clearly established.

For healthcare applications of artificial intelligence (AI), even the nascent case law that has begun for automobiles and self-driving cars is not firmly established. Although self-driving car systems are being tested on a regular basis at scale in the real world, AI systems for healthcare applications do not have the same authority over the lives of patients.

In the healthcare sector, what is broadly called AI takes a number of forms, from simple algorithms programmed to create efficiencies to machine-learning systems used to analyze images and make treatment recommendations. But some AI applications go beyond diagnostic tools, equipping robots to deliver treatment. As AI systems integrate with patient care and make more autonomous decisions, liability questions become more important in an environment without much legal guidance.

Healthcare Diagnostics and Tools Are Not “True AI”

“We’re not operating in the world of true AI yet,” said Tracey Freed, JD, a transactional attorney who teaches at Loyola Law School’s Cybersecurity & Data Privacy Law program. “We haven’t reached artificial general intelligence, where machines are making autonomous decisions on their own.”

To anticipate that development requires “thinking of what that world could look like and what liability looks like,” Freed said. “From a legal perspective, [AI could be considered] ‘agents’ of these companies or hospitals. Having these machines be my agent means I take on the liability for any incident that arises because of the machine. That’s the situation we’re more likely in today.”

An Israeli technology company, Ibex Medical Analytics, designed its Second Read system to check clinical pathology diagnoses. To train the system, the company fed it thousands of images of prostate core needle biopsies to teach it how to distinguish between benign and cancerous diagnoses.

Such tools—machine learning systems that are fed massive amounts of imaging or other data and trained to distinguish between different results&mdash;have become increasingly common. The federal National Institute of Standards and Technology has begun to push for the adoption of standards for measurements and medical imaging. These benchmarks could ease the design of machine learning systems and make them more broadly useful.

Determining Who Is at Fault When Things Go Wrong

Healthcare applications of AI occupy a critical position at the nexus of several important concerns for providers and patients, like privacy and human life, according to Lucas Bento, MS, an associate at Quinn Emanuel Urquhart & Sullivan, a large international law firm. Bento advises companies on the development stages of AI applications and technologies.

“The biggest question,” Bento said, “is how to allocate responsibility when the chain of causation is breached by, say, unexpected software error or hacking. Some theories of strict liability may be applicable.”

Strict liability is a legal claim that does not require a particular finding of fault, like negligence or intention. Product liability draws on this principle.

“In other instances, a court may take a deeper dive into what or who actually caused the error,” Bento said. “These are all novel issues that, barring legislative intervention, will be resolved via successive litigations across the country.”

But if litigation for these liability cases does not reach court, as in Herzberg’s case in the March Uber crash, the question of how to adjudicate liability for errors or injury involving AI cannot be established.

There have been other fatalities with AI systems involved in which drivers were killed, unlike Herzberg’s death in Arizona. Tesla’s Autopilot driving system was faulted by the NTSB for a fatal crash in Florida in 2016. The family of the deceased driver and their lawyer ultimately released a statement clearing Tesla of responsibility, but they refused to comment on any settlement.

READ: How AI Is Shaking Up Healthcare, Beyond Diagnostics

On March 23, shortly after Herzberg’s death, Walter Huang died in a fiery crash while driving a Tesla Model X and using its Autopilot system. B. Mark Fong of Minami Tamaki, the law firm representing Huang’s family, claims Tesla’s Autopilot feature is defective and caused the driver’s death. The firm is exploring a wrongful-death lawsuit on grounds that include product liability and defective product design.

Tesla has argued that its Autopilot system is not self-driving and that its user agreement requires the driver’s hands to be on the wheel when the system is engaged.

A. Michael Froomkin, JD, a law professor at the University of Miami and the author of 

, calls the human in the AI system the “moral crumple zone.”

“If there’s no human in the loop, it’s not controversial that whoever designed the tool is liable,” Froomkin said. “The AI [portion of the system] doesn’t impose anything special on that. ...The special bit is where the human is in the loop.”

Allocating responsibility between a human and AI in a system that relies on both parties is an unsettled and controversial issue, he said.

Designers of AI systems can claim they were not the last in the chain of responsibility, that AI is only advisory, and that the human is the decision maker.

Developers might also argue that a doctor or insurance company bears ultimate responsibility for negative outcomes or damages.

“If you have a human in the loop, if the system is advising a person, the person is going to take the fall,” Froomkin said.

As he put it, when linking this question to healthcare applications of AI: “When is it appropriate to blame the doctor using the AI? The easy case is where there’s no person. The harder case is when there’s a person between [the AI and the patient].”

AI doesn’t have legal standing because it does not have sentience. But “if [AI learns] on the job, you have an interesting liability problem: Was it taught improperly? Those are hard questions that are very fact dependent.”

According to Bento, product liability theories seem well suited to address questions of liability arising from the use of AI products and services. “Some challenges exist as to attribution of liability due to computer code errors,” Bento said.

Both Bento and Freed brought up other issues in expanding the use of AI, such as bias, and their implications for liability.

“Bias is another big issue for liability,” Bento said. “Eligibility for products and services is increasingly dictated by algorithms. For example, some consumer finance companies run algorithms to decide who is eligible for a loan or other financial product. Biased outcomes could create litigation exposure.”

Freed raised the issue of using algorithms in criminal sentencing to predict recidivism. In 2016, the nonprofit news organization ProPublica released a detailed investigation into how machine-driven “risk assessments” used to inform judges in sentencing showed bias against black defendants. One company, Northpointe, which created a widely used assessment called COMPAS, does not disclose the calculation methods used to arrive at final results. (Northpointe disputes ProPublica’s report.)

“Especially in healthcare, we need to know why the decisions are being made the way they are,” Freed said. “We need to know it’s the right cause rather than something that’s more efficient or something that’s the right call but for the wrong reasons.”

AI and machine-learning systems use data sets, which require careful thought and vigilance in their own right, Freed said.

READ: Holding Public Algorithms Accountable

SOPHiA Genetics and its SOPHiA AI genetics testing system, with its 200,000 patients tested, has a large data pool at this point to draw from and refine its assessments and predictions. Its international user base also gives it geographic range. Other machine-learning applications applied recently to analyze images must rely on a relatively small data pool to train the system’s clinical predictions.

“What’s interesting for healthcare [is that] a lot of what we’re doing is based on training data,” Freed said. “It seems [as if] your algorithms are only as good as your training data. A lot of the training data, if you don’t have a ton of resources, are coming from sources that are easily obtainable or free or not great sources of data.”

Recently developed machine-learning applications built to analyze medical imaging often rely on a relatively small data pool to train the system’s clinical predictions.

“What’s interesting for healthcare [is that] a lot of what we’re doing is based on training data,” Freed said. “It seems [as if] your algorithms are only as good as your training data. A lot of the training data, if you don’t have a ton of resources, are coming from sources that are easily obtainable or free or not great sources of data.” 

Cautious, Informed Optimism for AI’s Future

The risks don’t necessarily outweigh the promise of AI advancement, but healthcare and society at large must use caution alongside technological innovation.

“We need to take an approach where we’re excited but also more thoughtful and not innovating without being cognizant of the consequences,” Freed said. “We must take into consideration how these things can make either poor decisions or decisions that don’t make sense.”

Bento is also optimistic about the benefits of AI systems despite the attendant new and uncertain legal risks.

“New ways of doing things always create risks of unintended consequences that may result in harm,” Bento said. “The beauty of AI, though, is that it doesn’t just provide a new way to do something but, if done right, a better way of doing it.”

Bento also argued that AI’s potential to reduce human error is an important factor that users should not ignore. That potential for error reduction is something that Tesla has strongly argued when its Autopilot system has been linked to fatalities.

After Huang’s fatal crash in Mountain View, California, the company pushed back against any claims that its Autopilot system was defective. In a press release, Tesla touted the ability of the system to drastically reduce the risk of any fatal accident, ultimately making roads safer, and it claimed that the system’s efficacy continues to improve.

The National Highway Traffic Safety Administration, which includes the NTSB, “found that even the early version of Tesla Autopilot resulted in 40 percent fewer crashes,” Tesla said.

Tesla also aggressively criticized the NTSB for chasing headlines and inappropriately focusing on “the safest cars in America” while “ignoring the cars that are the least safe.”

Some auto manufacturers working on self-driving technology have pledged to accept full liability for accidents involving their self-driving cars because of error, including BMW, Mercedes-Benz, and Volvo. The companies are betting that with more testing and wider implementation, the number of accidents will plummet.

Articles

Healthcare can learn from self-driving cars’ fatal lessons in liability.

Who Is Responsible When AI Fails?

Image has been resized and cropped. Courtesy of 

Greenphire, a financial software company, announced a partnership with Lyft this week to centralize billing for the transportation of clinical trial participants, along with other payments, furthering the idea that Silicon Valley-style disruption is gripping healthcare.

The partnership would allow users of Greenphire’s ClinCard service to order a Lyft ride and provide ride reimbursement. ClinCard automates payment to participants in clinical studies, allowing site coordinators to disburse funds to a reloadable debit card, according to Greenphire’s website.

READ: Jeff Bezos’s VC Arm Contributes to Mindstrong’s $15M Round

Greenphire’s CEO Jim Murphy said in a statement that Clincard is “the only application that allows a study coordinator to order a car, reimburse for a ride, and provide a stipend all within one software.”

The company’s press release touts the partnership with Lyft as one that will allow centralization and increased efficiency, increase financial transparency, and make trial participation easier for patients.

Gyre Renwick, vice president of Lyft Business, was quoted in the press release saying Lyft’s commitment to accessibility and push to eliminate complications for patients.

Greenphire frames ClinCard as a fast method for reimbursement that helps keep patients in clinical trials by giving them quick access to cash and automated reminder messages. Partnering with Lyft to allow site coordinators to provide transportation to patients could further smooth patient access and retention and smooth patient access.

But a study by the Perelman School of Medicine at the University of Pennsylvania 

 about just how much Lyft’s ride service increases the likelihood of patients making appointments.

The study, which tracked 786 Medicaid recipients in West Philadelphia, showed virtually no difference in the rate of missed appointments between those who accepted the offer of a free Lyft ride and those who did not. And only about 30 percent of those patients who were offered the free ride accepted the service.

Integrating Lyft ride scheduling within ClinCard may streamline payments, notifications to patient participants, and scheduling within a pre-existing system that already serves over 1 million trial participants, according to Greenphire. But based on the Perelman School study, Lyft ride scheduling might not increase the likelihood of patients to make their appointments.

Lyft isn’t the only ridesharing service to enter the arena of patient transportation. In March, 

, which bills healthcare providers and provides a system to schedule rides for patients, caregivers, and staffers. The division could also aid telehealth patients in retrieving prescriptions, 

according to Uber Health head strategist Lauren Steingold

Get the best insights in healthcare analytics 

A Free Ride With Lyft Didn't Reduce Skipped Medical Appointments

Which Health-Tech Start-Ups Are Making Money in 2018?

AbleTo Is Changing How We Use Behavioral Health Data

The rideshare company’s new partnership is another step toward tech disruption in the industry.

Lyft Dives Deeper into Healthcare

The sudden disruption hit electronic medical records (EMRs) provider AllScripts sometime between 2 AM and 6 AM on January 18. Two of the provider’s data centers had been targeted by a ransomware attack, locking up Allscripts-managed data in an extortion attempt. About 1500 of its clients—mostly small practices, according to the company&mdash;were unable to access their records for 8 days.

Like many ransomware attackers, the Allscripts hackers offered to unlock their victims’ data in return for a ransom payment in bitcoin. The electronic cryptocurrency is popular with hackers because it is virtually impossible to tie the digital currency to its holder. It’s their preferred method of payment.

Big Healthcare Players to Explore Blockchain's Viability

But blockchain, the very technology that underpins bitcoin, might have applications that help secure data from malicious hackers and others who seek to exploit private data, like EMRs and patient information. At the same time, blockchain has burst onto healthcare’s radar to great applause, much of which might be premature or unfounded.

A Secure Blockchain Network for Personal Data

One possible application is Datum, a blockchain network in development, whose cofounder and CEO Roger Haenni describes it as “eBay for data,” a secure marketplace where individuals would control the use of their private information and profit from its sale.

When Haenni first conceptualized what would become the Datum network, he was busy developing something comparatively prosaic: Clever Baby, a wearable device much like a fitness tracker, would be used to track an infant’s vitals. Haenni’s team took the device to the prototype stage and set about planning practical trials. But in conversations with hospitals to develop trials of the Clever Baby prototype, Haenni kept running into pushback.

“We always ended up with an ethics board at the hospitals we went to,” he said. “This whole discussion started about putting the device on babies. What data is collected? Is that compliant with the data collection and privacy laws we have in our jurisdiction?”

According to Haenni, the cautiousness around data collection stemmed from liability concerns. “All of these discussions were focused around the risk [aversion] of the hospitals,” he said. “They don’t care about the parents or babies so much, but rather about liability.”

Concerns regarding patient data collection and the responsibility to protect that data aren’t trivial. Due to Health Insurance Portability and Accountability Act (HIPAA) reporting requirements, the Department of Health and Human Services maintains a public database of breaches. So far in 2018, the department lists 219 breaches due to “unauthorized access” or “hacking incidents.” Allscripts is currently subject to a class action suit brought by affected clients who allege negligence and breach of the company’s duties under HIPAA to protect their health records.

These conversations about compliance and data protection prompted Haenni to think about how to enable everyone to secure the vast volumes of data produced by everyday activities and select who is able to access and use that information. “What is the real large issue here that we can solve?” Haenni said, describing the question Datum sought to answer.

It was “this notion of returning data ownership and giving back control to the data creators who are the individuals in many cases,” he added. “That was something much more important that would have a larger impact on the world than baby wearables.”

Blockchain technology provided the solution.

Blockchain is often described as a “distributed public ledger,” a public record of all transactions across a network, protected from tampering through encryption. Cryptocurrencies like bitcoin combined existing technologies—for example, public and private key encryption&mdash;in a novel way to answer straightforward questions: What is the best way to secure digital currency against malicious attacks or duplication? What about exchanging value in the absence of trust or central authorities like national governments?

Transactions across the network are packaged into blocks. Special members of the network, called miners, use computing power to solve a difficult mathematical problem to validate the block, definitively linking it with the rest of the chain, which is secured by encryption. In cryptocurrency networks, the member who first validates the new block is rewarded with the network’s currency.

Newer blockchain applications extend this mechanism to accomplish tasks far afield from currency transactions. Social network Steemit, for instance, rewards users for publishing popular content; Civil is a platform for blockchain-based journalism that runs on the Ethereum network.

The essential function of blockchain, argued attorney Steven W. Teppler, JD, is provenance. Teppler leads the Abbott Law Group’s technology litigation practice. He recently participated in a webinar with 

™ to explore the possibilities for blockchain applications in healthcare.

“It’s a transparency mechanism on steroids. It would expose fraud,” Teppler said. “What blockchain does is provide an audit trail—no more, no less.”

In the case of a blockchain used to encode healthcare data, “someone can carry around a key to their entire history and not fill out a new one for each provider they go to,” Teppler said.

“Every data instantiation is signed by the provider, and that digital signature, in turn, [is] timestamped so that both the signature and the data can be vetted years down the pike.”

It’s an often-unquestioned bargain that to use free services like Google’s search or email services or Facebook, users supply vast amounts of data on a wide range of activities. Companies then use that data almost without restriction and in hushed ways. The huge amount of data collected by Google and Facebook is available to advertisers to better target ads, and Google constantly collects phone location data to calculate information, like traffic speed and density, for Google Maps.

Data providers, however, go uncompensated—unless you count free access to these services. Once that data are collected, its use and resale is largely out of an individual’s control. Datum allows a user to decide what information to share, who to share it with, and the terms of use. Any information a user opts to share or store is encrypted and uploaded to a decentralized database made up of members of the network, with a transaction recorded on Datum’s blockchain. Buyers pay users in Datum’s currency, called DAT.

Although Haenni said Datum will initially focus on connecting users with data to sell with advertisers looking for information to collect, he envisions healthcare data as a possible future use. “[Health] is just one of the verticals that can be built on our infrastructure,” Haenni said. “We’re looking to partner with other companies who have more realm knowledge—other developers who will build their services on top of [Datum].”

Haenni contrasted Datum with Apple’s HealthKit offering, which allows developers to leverage health data collected by Apple devices for new apps and services. But it’s unclear, according to reporting by The Outline, whether those data are protected by HIPAA. That means the same ethical concerns Haenni’s partners had about Clever Baby’s data collection are also present for HealthKit. How developers may end up using, distributing, or reselling HealthKit data is unknown.

“Datum uses blockchain to provide a public ledger to, in turn, provide transparency of who accessed what data at what time, whether there was a monetary transaction, or what this data was pulled for,” Haenni said. “We don’t want to lock this data away but to unlock this data under much more transparency and control. If an advertiser uses [your data], you should be compensated, you should know about it, and you should have the ability to refuse.” Individuals would also have the ability to make their data freely available to trusted organizations, like research institutions, he added.

Datum’s design raises the possibility that an individual’s healthcare data, like EMRs and patient health information, could be securely stored on its networked database, under the full control of an individual. With data made anonymous, encrypted, and distributed over a global network, attackers might not be able to unlock and abuse any data they manage to steal. 

The rapid proliferation of blockchain startups and the checkered history of trading cryptocurrency makes figuring out how to usefully apply the technology bewildering. Bitcoin, the first blockchain implementation, has scored headlines for its extreme volatility. The exchange rate rocketed from just under $1000 at the start of 2017 to nearly $20,000 by the end of the year.

Cryptocurrency trading exchanges are periodically hacked, suffering eye-popping losses. In 2014, early heavyweight Mt. Gox lost around 850,000 bitcoin, valued at close to $450 million at the time ($5.7 billion at the time of publication). That record was surpassed this year when Japanese exchange Coincheck reported a loss valued at approximately $533 million.

Matt Suiche, the founder and director of cybersecurity startup Comae Technologies, is skeptical about the current development stage of blockchain. “What we see that’s popular now may not be what’s popular in 4 years,” Suiche said. “[Blockchain] technologies are not mature yet. They’re still experimental. But people treat them like they’re mature and will solve all their problems…Need a database? Add a blockchain. Need something secure? Add a blockchain.”

But security, Suiche pointed out, is a complex problem. Although the blockchain itself might be immutable, thieves can still prey upon a blockchain network. Some of their techniques are well-known classics, like phishing. In one case from July 2017, Israeli startup CoinDash lost a large portion of its initial coin offering, the equivalent of about $10 million at the time. Hackers changed CoinDash’s website by substituting their own address to receive investment funds for the address belonging to CoinDash.

Succeeding blockchain networks have evolved beyond some limitations of bitcoin. Ethereum, perhaps the most popular bitcoin successor, enables the use of “smart contracts”—bits of code, like computer programs&mdash;on its network, published to the blockchain and executed by members of the network.

Ethereum has a host of heavyweight backers, eager to take the decentralized public technology and apply it to private enterprise. The Enterprise Ethereum Alliance, which aims to develop standards and research enterprise applications for the network, counts among its members Deloitte, Accenture, JP Morgan, Credit Suisse, Intel, Microsoft, and Pfizer. Some companies have adapted the open source Ethereum code to build specialty enterprise blockchain networks; for example, Quorum, JP Morgan’s implementation of Ethereum.

Smart contracts allow a dizzying array of distributed applications, built on top of the Ethereum network, used by Datum. Ethereum’s blockchain offers more than a digital currency, although it produces a cryptocurrency called ether.

Smart contracts have their virtues. Cryptographically encoded into the near-inviolable blockchain, they can’t be tampered with. But like any other program, they are vulnerable to user error.

In 2016, an early Ethereum application run by smart contracts threw the network into crisis. The Distributed Autonomous Organization (DAO), an automated crowdfunding network run on Ethereum, lost the equivalent of around $56 million in ether to an unknown attacker who exploited a bug in the DAO’s smart contract design. In response, Ethereum’s developers and proponents of the DAO mounted a successful campaign for a “hard fork,” in which the entire network would be duplicated in a state where the theft never happened, essentially erasing history. This kind of collusion is what blockchain is supposed to prevent, yet most users cosigned this effort, shifting over to the hackless fork and abandoning the version where the hack had taken place.

Private blockchains, the kind that might be run by a group of participating companies, could be even less resistant to collusion. Private blockchains have a limited number of influential participants compared to the vast number who could potentially participate in public blockchains like bitcoin or Ethereum. A private blockchain could potentially be rewound or rewritten if its participants collaborate to alter it, obviating the technology’s value as an inviolable ledger.

Suiche developed a tool called Porosity to take apart and examine smart contracts already operating on the Ethereum blockchain. Suiche said that JP Morgan integrated Porosity into its Ethereum-based Quorum blockchain. A simpler smart contract design is safer, Suiche said, and has less chance of including unnoticed vulnerabilities. But one of the core virtues of smart contracts—their immutability when locked into the blockchain&mdash;is also a liability.

“Smart contracts are just like any other piece of software,” Suiche wrote in an email, describing what he learned from implementing Porosity. “They can have vulnerabilities, and the languages designed for smart contracts will need to be [mathematically] formally verifiable to reduce the number of issues.”

Asked about Datum’s use of smart contracts, Roger Haenni compared the risks of smart contracts to the risks of blockchain technology itself. “With bitcoin, you are putting your trust into the code that runs the bitcoin network,” he noted. “No one can alter this code without a majority of miners agreeing.”

The kind of alteration Haenni is talking about is similar to the change made by the Ethereum network to rewind the effects of the DAO hack. Even that, however, took a developed consensus arrived at through community discussion, not the arbitrary interference of a central authority.

“Your assets on the blockchain are protected by the impartial and hopefully bug-free code without the possibility of any central party interfering,” Haenni wrote. “It is precisely this immutability of the code that is critical to give everyone trust that the system cannot be gamed. But like any piece of software there may be bugs.

“The key is that the core systems and code have been reviewed many times by a huge community of developers and, ultimately, that is what you have to trust. With smart contracts, the problem is that anyone can write a smart contract, and you have to trust the creator and the team behind it.”

Suiche added that companies considering implementing blockchain take a step back and carefully consider their actual needs. Nonblockchain solutions might be easier and safer to address a particular need.

“Everyone is going to have a different use case,” Suiche said. Companies need to “identify what’s important for them and not focus on the technology, but rather on key verticals that they need to invest in.”

The technology appears capable of safeguarding sensitive patient data, but it has yet to make a big footprint outside the cryptocurrency community.

Is Blockchain the Answer to Healthcare's Cybersecurity Concerns?

The first tweet that the Dark Overlord sent to a small chiropractor in Poughkeepsie, New York, read: “We’re watching you. Make the right choise [sic].”

By the next day, it was too late for choices. The hacker group had publicly named the practice and claimed to possess all its patient information. The chiropractor, the hackers said, had “rejected our most handsome proposition.”

The pattern is typical of the Dark Overlord: Choose a vulnerable target, steal its data, announce the theft, and demand payment via Bitcoin. If the victim does not pay, the group threatens to release the information or sell it on dark web exchanges, the anonymous underbelly of the internet.

A mythology has come to surround the group, which perpetuates the lore, sometimes by tweeting ominous religious passages. In one extortion letter, it claimed responsibility for “some of the most serious breaches and security violations in the last year.”

The Dark Overlord punches above its weight through strong branding, a focus on terrorizing its victims, and a deliberate press strategy. Although relatively small compared to the $4 billion in global damages wreaked by the WannaCry ransomware attack, its successes show why healthcare organizations of every size must ready their cyber defenses.

Why the Dark Overlord Wants Healthcare Data

The electonic theft of medical records, including health and financial data, appeals to hackers. Among all industries, healthcare suffered the second largest number of data breaches in 2016, according to Symantec’s Internet Security Threat Report. (Strict reporting rules may have affected the ranking.) Last year, healthcare organizations reported roughly 130 breaches that burned as many as 3.3 million people.

The average cost of a data breach is $7.4 million, including customer loss and the costs of notification and remediation, according to an international study of 419 companies across all sectors. Medical information is especially attractive to thieves, experts said, because of its long shelf life and high value. “The chance of being able to run fraudulent activity around them is higher,” said Waylon Krush, CISSP, CISA, who is CEO of the security company Lunarline.

Patient records fetch better prices than other personal information does, experts noted. It is “high payoff,” Krush said. “[Hackers can] try to get to prescription drugs or trick the government into paying for different healthcare activity may not have happened.” He mentioned a long-running, lucrative Medicare scam in which fraudsters used member information to get the government to pay for unnecessary powered wheelchairs.

Money-making schemes go beyond fraudulent billing, said Stan Banash, MBA, chief information security officer at Children’s Hospital of Orange County in Orange, California. “If [a hacker] has medical records for famous people, you can sell those because there’s always someone willing to buy dirt—medical conditions people find embarrassing and don’t want public. There’s a possibility for extortion,” he said.

If a credit card number is stolen, Banash pointed out, credit card companies and banks can kill the card and issue a new one. “It’s not the same with a medical record. If you have a genetic condition, you can’t change that,” he said.

The person or person behind the Dark Overlord are unknown, but its victims sometimes tend toward the higher profile. They include Hollywood production studios and public school districts—and healthcare institutions, on which the group first cut its teeth. Its targets range in size from a California eye doctor to large clinics and insurers.

™ attempted to contact several targets but did not receive responses.

The Dark Overlord reached national notoriety in December 2016, when the group broke into the server of a Hollywood postproduction studio and nabbed 10 unreleased episodes of the Netflix series 

. Hackers demanded $50,000 in Bitcoin for their return. The group’s largest healthcare job, meanwhile, may be its haul of 9.3 million patient records from what it called “a large insurance healthcare organization in the United States.” After the company refused to pay, the Dark Overlord offered the records for sale on a dark web marketplace.

To pressure healthcare hacking victims, the Dark Overlord has promised to embarrass famous patients. After attacking a London-based plastic surgeon, the group threatened to release before and after photos of patients, which allegedly included British royals and celebrities. In November 2016, the group threatened to release the records of college and professional athletes who attended an orthopedic clinic in Atlanta, Georgia.

The Dark Overlord’s method of getting payouts for patient data is more personal than that of other cyberattackers. In a ransomware attack, a malicious actor breaks into a network and locks the victim’s data, and then offers a key to unlock it—in exchange for payment. Interaction between attackers and targets can be as brief as transmitting the ransom and, depending on the reliability of the hacker, receiving a decryption code.

Unlike the Dark Overlord’s victims, who seem to be intentional, ransomware targets are often random. In May 2017, the WannaCry ransomware attack affected 200,000 computers across 150 countries, temporarily crippling the United Kingdom’s National Health Service and resulting in an estimated $4 billion in losses. WannaCry was self-propagating, spreading via email as viral malware. The attack probably was not intended to target large institutions, considering the Bitcoin ransom price was about $300, according to experts.

Victims know immediately when they have fallen prey to ransomware. They find their systems locked and unusable, their screens displaying a countdown clock and an address accepting Bitcoin payments. In contrast, the Dark Overlord’s victims may not realize they have been hacked until they receive a taunting email, like this extortion note sent to one school district: “If you receive a message from us, it means you have been completely and thoroughly attacked and breached by an organised entity of creatures who are motivated only by their love for internet money,” the Dark Overlord wrote. “We are savage creatures who do not discriminate. We prefer to prey upon the likes of institutions such as your own, but not because we have anything against children, but rather for much more interesting reasons which you will soon come to understand.”

Hackers then reportedly used the seized data to send threatening text messages to parents and students, forcing the temporary closure of 8 schools.

Ryan Kazanciyan, chief security officer at the cybersecurity firm Tanium, called the Dark Overlord’s extortion attempts “less common than classic ransomware campaign or other forms of extortion, like denial of service attacks.” The Dark Overlord’s direct approach involves more unknown factors. “How do we know [the attacker has] the data? How do they know we’ll comply with demands?” Kazanciyan said. “When you think of the economics of cybercrime, you need consistency.”

Healthcare must bolster its defenses against all cyberattacks, including those from the Dark Overlord. The spike in the rate of healthcare hacks, however, might be due not to poor industry standards but, rather, to greater security awareness. “The increased popularity of ransomware has made breaches more visible,” Kazanciyan said. “Stealing and selling data on the black market can sometimes be silent.”

Krush echoed Kazanciyan’s conclusions: “Until fairly recently, most healthcare organizations didn’t even know if they experienced a breach. They weren’t investing in security and had no response capability.” Now, institutions know when they have been hit.

Healthcare must first focus its everyday efforts on solving the problem, according to Banash—and he should know. “We get attacked on a daily basis,” he said. “Our network is being probed a dozen times or more a day by people doing reconnaissance.” Aging, unsecured medical devices offer hackers a particularly dangerous window through which they can jump to other locations on the same network. Those devices can also, if locked with a ransomware attack, be crippled and unavailable for patient treatment, which happened to the British National Health Service during the WannaCry attack. Hospitals must focus on protecting these high-cost, heavy-duty devices, experts said.

In this age of consolidation, healthcare organizations may also benefit from inspecting their systems before linking arms. “One company can inherit malware and existing breaches from the other,” Kazanciyan said. “In other cases, you’re taking 2 mismanaged, insecure environments and combining them, creating something that’s even more insecure than the sum of its parts.”

The first step to strengthen cybersecurity? “An accurate and up-to-date inventory,” according to Kazanciyan. Tech specialists must maintain computing devices, operating systems, and software, keeping them updated and patched. Both the Dark Overlord and WannaCry attacks took advantage of vulnerabilities in older operating systems that their owners had not updated or addressed.

But before anything, healthcare groups must realign their budgets to support their cyber defenses. “The mind-set has been [that] every dollar put into security is being taken away from patient care, but there is a line where underinvesting in security can have a real effect on patient care,” Kazanciyan said. “Security isn’t just to check a HIPAA compliance box; it’s also to prevent catastrophic effects on patient care.”

Although the Dark Overlord's targets tend to be small practices, large organizations face the same threat and corresponding challenges, experts say. The hacking group typically exploits long-standing vulnerabilities in old systems, and more of these may exist in a bigger institution.

“Ten times the number of computers means hundreds of times the levels of risk,” Kazanciyan said. “It’s basic system management: knowing what computers are on the network, what software they’re running, how they’re allowed to communicate with each other. With large healthcare orgs, that doesn’t necessarily go away.”

And the danger posed by the Dark Overlord and any number of similar hackers is not going away, either.

No one knows who is behind the hacker collective, but hospitals must be prepared to fight these self-described “savage creatures.”

Defending Your Data From the Dark Overlord

A pharma giant just completed a major programmatic ad buy, aiming to reach a rarefied audience: the small population of American oncologists. Through the exchange, the ads land on Oncology Tomorrow, a brand-new and fast-growing website. After a month, Google Analytics shows, the website had gathered more than 100,000 visitors, a similar number of search referrals, and 250,000 page views.

But virtually all of Oncology Tomorrow’s traffic is fake. A follow-up analysis estimates that 0.03% of the site’s impressions were human—only 76 of roughly 220,000&mdash;with bots making up the rest. These are the automated programs that simulate ad impressions, click-throughs, and even profiles of high-value visitors like the oncologists targeted by this fictional pharma company. 

Oncology Tomorrow, however, is real. Yet it is not the work of a malicious hacker. Wei Han Frank Lin, chief technology officer at DMD Marketing, built the website and its army of bots. Now he points to the experiment as a compelling demonstration of how easy it is for healthcare companies to lose advertising dollars to high-tech ad fraud. “When we looked at the actual data, only 1 physician had ever visited the fake experimental site,” he said.

The scale of ad fraud is striking. The third annual ad fraud report from the Association of National Advertisers and the cybersecurity group White Ops shows an estimated $6.5 billion lost to the practice in 2017, down from an estimated $7.2 billion the year before. The study used specially designed detection and tracking software in partnership with 49 major companies, including Unilever, Walmart, and AT&T. The report attributes the drop to increased awareness and better detection techniques from individual companies, but the overall decline in digital ad fraud appears slight.

Oncology Tomorrow and the many sites like it are surprisingly difficult to detect. But with millions of dollars at stake, it is critical that decision makers across the healthcare field, from pharma to hospitals, understand the nature of this threat.

At first glance, Oncology Tomorrow is indistinguishable from other medical information sites. There is nothing obviously fraudulent about it. It is well designed, with a fresh logo, catchy headlines, and colorful stock photos. Click a link, and visitors get a short blurb describing, for example, a recent Drexel University study showing that an increase in cigarette prices will lead to more smokers quitting.

But it is a front, a ghost site not meant for human consumption. Any human visitors are unintended side effects. The site is an automated advertising money drain, crawling with Lin’s bots. Companies pay the rates for cost per click and cost per thousand impressions (CPM) but without actually reaching the humans they believe they are.

Lin used multiple methods to load Oncology Tomorrow with bots. He harnessed traffic brokers, for which low up-front costs almost ensure that all subsequent activity will come from bots. Lin also used “click farms,” an arrangement in which human workers, often in third world countries where wages are low, click through sites over and over again. And Lin purchased bots from “the darker side of the Web” and customized them.

“All websites love search term referral lending,” Lin said. “I created a bot that pretends to be a human being, submits a search for Oncology Tomorrow, clicks through on a [search] result, pretends to have mouse movement, and clicks on ads as a click-through.”

Bots can also simulate demographic profiles used to target ads, said Augustine Fou, a fraud and cybersecurity researcher with Marketing Science Consulting Group. By visiting sites that physicians might visit and picking up cookies, bots camouflage themselves with a browsing profile identical to those of actual healthcare professionals.

Bots, for instance, can visit the sites of prestigious medical journals. Based on a broad look at these medical publishers, Fou believes bot traffic to legitimate journals is small. “They go there first, collect a cookie, and it looks like they visited the journal, and then [the bots] go elsewhere and attract retargeting dollars,” he said. Companies “pay more money to try to retarget them when they show up elsewhere. Bots are making more money by posing as a valuable visitor and going elsewhere to cause the impression.”

Or those bots collect cookies and use their profiles on traps like Oncology Tomorrow, what Fou called “fake cash-out sites.” Past reports suggest that bots and these honeypots have invaded the digital ad economy. And crawlers even earn higher click-through rates than humans.

Another common ad fraud technique is “spoofing,” in which fraudulent publishers imitate the URLs of prestigious sites, using bots to simulate impressions and clicks. In December 2016, White Ops revealed a criminal bot network it dubbed Methbot, named for references to the drug meth in its code. The bot network was earning $3 million to $5 million per day. It spoofed more than 6000 “premium domains,” according to the company. The ads were real, but only bots were watching.

 discovered that it was one of those premium publishers targeted by spoofers. Fraudulent ad inventory purporting to be from the 

, according to a report, was worth $1.3 million per month. 

Pharma and other sectors of healthcare might face a greater challenge than the rest of digital marketing. This is because of higher ad rates paid to reach a more specialized audience, Lin and Fou said.

To combat this, Fou suggested, companies should be wary of large numbers. “It’s trivial for bots to pretend to be whomever you want to see,” he said. When trying to reach oncologists, for instance, it is important to remember the small size of the desired audience. There are around 15,000 oncologists or hematologists and another 8500 subspecialists, according to American Medical Association listings. How could a legitimate site attract that total, let alone a larger audience?

It is also key to search for “look-alike” audiences, Fou said: “They see if other users visit the same set of sites. If they did, we assume they are the same type of user.”

The solution offered by Lin’s company, DMD Marketing, is an opt-in white list, a way to sidestep the technology arms race between bots and detection technology. The audience identity management platform “will allow us to authenticate and identify a physician in real time,” Lin said.

For smaller audience targets, ad exchanges and publishers deserve high CPMs, Lin said. “How they are measured is the problem we think they need to fix. If the industry can do a deterministic way to identify actual targeted impression, we would advocate even higher pricing for that.”

Fou, meanwhile, recommended that pharma and hospital marketers optimize for business outcomes to minimize risk. “Don’t go after reach and frequency. This is not a branding game,” Lin said. “It’s about niche and how tight your targeting is going to be for pharma marketers.” Asking for quantity where it doesn’t exist opens the door for fraud. To avoid being suckered by spoofed URLs on ad exchanges, he suggested replacing programmatic ad buys with direct buys and demanding transparency and more data from advertising agencies.

In the end, this issue is not about fearmongering but vigilance. “The good news,” Lin said, “is that publishers, and especially reputable partners, are sincerely trying to get the best outcome for patients and the physicians.” The bad news is that hospitals and healthcare at large must continue to keep an eye out.

A pharma executive launched a “cash-out” site to reap ad dollars from companies targeting high-value physicians. How can healthcare protect itself?
 