`
`Markus Jakobsson
`
`1 My Background
`• Focus. Identification of security problems, trends and solution along four
`axes – computational, structural, physical and social; quantitative and
`qualitative fraud analysis; development of disruptive security technologies.
`• Education. PhD (Computer Science/Cryptography, University of Cali-
`fornia at San Diego, 1997); MSc (Computer Engineering, Lund Institute
`of Technology, Sweden, 1994).
`• Research labs. San Diego Supercomputer Center (Researcher, 1996-
`1997); Bell Labs (Member of Technical Staff, 1997-2001); RSA Labs (Prin-
`cipal Research Scientist, 2001-2004); Xerox PARC (Principal Scientist,
`2008-2010); PayPal (Principal Scientist, 2011-2013); Qualcomm (Senior
`Director, 2013-2015); Agari (Chief Scientist, 2016-2018) ; Amber Solutions
`(Chief of Security and Data Analytics, 2018-)
`• Academia. New York University (Adjunct Associate Professor, 2002-
`2004); Indiana University (Associate Professor & Associate Director, 2004-
`2008; Adjunct Associate Professor, 2008-2010).
`• Entrepreneurial activity. RavenWhite (Authentication solutions; Founder,
`2005); Extricatus (Security consulting; Founder, 2007); FatSkunk (Mal-
`ware detection; Founder, 2009); LifeLock (Id theft protection; Member of
`fraud advisory board, 2009-2010); CellFony (Mobile security; Member of
`advisory board, 2009-2010).
`• Anti-fraud consulting. KommuneData [Danish govt. entity] (1996);
`J.P. Morgan Chase (2006-2007); PayPal (2007-2010); Boku (2009-2010);
`Western Union (2009-2010).
`• Intellectual Property. Inventor of 100+ patents; expert witness in sev-
`eral patent litigation cases (McDermott, Will & Emery; Bereskin & Parr;
`WilmerHale).
`• Publications. Two books – Phishing and Countermeasures: Understand-
`ing the Increasing Problem of Electronic Identity Theft (Wiley, 2006);
`Crimeware: Understanding New Attacks and Defenses (Symantec Press,
`2008); 100+ peer-reviewed publications (list available separately).
`
`1
`
`USR Exhibit 2102, Page 1
`
`
`
`2 My Beliefs
`
`What is security? Computer security allows the enforcement of rules, stan-
`dards and desirable behaviors. It hinges on an understanding of how systems
`can fail, followed by a design – whether on the protocol level or algorithm level –
`that forces a potential adversary to align his behavior with the desired behavior.
`Systems that achieve this are secure.
`However, it is not meaningful to speak about security without first under-
`standing the goals and limitations of the adversary: security is a relative concept.
`A system can be secure against one form of abuse without being secure against
`another. This makes it crucial to understand the constraints of potential ad-
`versaries. These constraints may be computational (requiring an understanding
`of lower bounds of computational efforts), structural (related to an understand-
`ing of protocols and business models), physical (e.g., be relative to how an
`adversary can manipulate devices), and social (such as in the context of phish-
`ing, where consumer psychology is defining the abilities of the adversary.) The
`client devices and the security measures running on these can be characterized
`in a similar manner. Computer security requires an understanding of all the
`dimensions of a problem, and the devices or resources we wish to protect.
`
`Computer security: invisible and dynamic. No matter where we look,
`abuse is conceivable. The failure to defend against it can affect the privacy
`of individuals, the intellectual property investments of organizations, and the
`integrity of governments. Everybody is affected. But not everybody should
`have to worry: Security should be automatic, natural, and built in.
`Since abuse is not static, but is a reaction to opportunities, security must be
`proactive. This makes it vital to understand trends – in computational ability,
`device deployment, application usage, and social and human constraints. To
`the extent that it is possible, these features must be measured, their results
`extrapolated.
`
`3 My Work
`
`How I select what to work on. My work is cross-disciplinary, and driven
`by market forces – before they express themselves.
`I spend time identifying
`potential trends and testing hypotheses before I embark on an effort. Let me
`give two examples to explain what I mean.
`• Social aspects of security. In year 2001, years before media used the
`term “phishing”, I was put in charge of identifying trends that could affect
`the revenue of my employer (RSA Security). I anticipated that the growth
`of online commerce would result in an increase in online fraud, and iden-
`tified it as a socio-technical computer security problem. (This belief was
`considered bordercase ludicrous at the time.) My work on phishing grew
`out of a conviction that this would become a significant issue. The needs to
`understand the human aspect of the problem led me to study psychology
`
`2
`
`USR Exhibit 2102, Page 2
`
`
`
`and to develop experimental approaches that can measure vulnerability to
`fraud (standard surveys will not). This fueled my work on ethical experi-
`mentation, and required an understanding of the goals of Internal Review
`Boards.
`
`Having identified what makes people vulnerable, I could establish techni-
`cal goals. For example, I had determined that the most common way for
`consumers to identify phishing attempts was by identifying the presence of
`incorrect information (as opposed to the absence of correct information).
`This begged the question whether a fraudster could construct an email
`message that will say ”Dear Wells Fargo client” only to actual Wells Fargo
`clients, and ”Dear PayPal user” to actual PayPal users? (My students and
`I established that the answer is yes; see browser-recon.info for a demo.)
`Similarly, I knew that password reset was – and still is – one of the weak
`links. This prompted me to develop alternative ways for backup authen-
`tication that did not rely on long-term memory; which avoided common
`attacks; and which were usable. (For a demo of one way to address this
`problem, see www.visual-blue-moon-authentication.com.)
`
`One of the attached papers (not yet published) describes how the
`security of this password reset scheme can be quantified.
`• Physical aspects of security. Similarly, my understanding of the power
`limitations of handheld devices has driven my work on alternative anti-
`virus paradigms for mobile computing. I am convinced that the status quo
`is doomed: One cannot constantly run a computationally intensive pro-
`cess on a device with limited power. (This is not a practical problem yet,
`as the power consumption depends on the number of malware threats in
`a manner that varies between logarithmic and linear, and there is hardly
`any mobile malware in circulation. In 2-3 years, as smartphones are esti-
`mated to surpass Windows machines in commonality, it will be a real and
`pressing problem, though.) This insight fueled my work on software-based
`attestation, where I have developed the first provably secure technology –
`no heuristics needed – for which the execution time to screen a device is
`a matter if milliseconds. This based on the simple fact that in order for a
`malware agent to be active, it needs to take up some space in RAM. If all
`RAM overwritten by a pseudorandom string, the malware agent needs to
`relocate some of this data; later, if we compute a keyed checksum of the
`entire RAM, then this relocated data needs to be accessed. This causes a
`delay, which then becomes indicative of the existence of an active malware
`agent.
`
`One of the attached papers (not yet published) describes this
`approach in detail.
`
`My work on mobile malware also motivated me to understand how carriers
`can use anomaly detection to identify and classify epidemics. Virus or
`trojan? Propagated by Bluetooth or SMS? It is possible to tell without
`ever analyzing the malware code, but just by looking at how it moves
`
`3
`
`USR Exhibit 2102, Page 3
`
`
`
`through the network. The detection is fast, central, and does not drain
`the batteries of the client devices.
`
`Management style. I have managed groups of researchers in a number of
`different roles – most recently as Principal Scientist at Xerox PARC, and as a
`Founder/CTO of FatSkunk, previously as a faculty member managing PhD stu-
`dents and junior colleagues. I believe in delegation and personal responsibility,
`and begin management relationships by determining the degree of independence
`(and desired independence) of each team members. I do not micro-manage, but
`I do provide suggestions for improvements.
`
`Vision of future needs. It is not meaningful to try to defend against a
`threat that one does not understand. The first step must be to understand and
`quantify the problem, and to recognize what constrains the possible solutions.
`This must be done in terms of the computational, structural, physical and social
`dimensions.
`There is a substantial need for work that secures the infrastructure, whether
`from technical or social threats. This will involve malware detection and recov-
`ery; robustness against denial of service and denigration attacks; establishment
`of identity (whether device or user); maintenance of trust (on both a technical
`and human level); user communication (including avoidance of social engineer-
`ing, how to communicate important information to unmotivated users, and how
`to build security mechanisms that are usable in the face of adversarial cam-
`paigns). There is also need to recover from failures on various levels; and to
`use anomaly detection for early-warning systems. It is important to understand
`that user behavior will change dramatically in situations of attack, and this may
`in itself destabilize systems. To address these issues, a broad understanding of
`vulnerabilities, technologies, and trends is necessary.
`
`4 Three Sample Publications
`
`I have published two books and 100+ peer-reviewed articles. An extended ver-
`sion of this document with three sample publications at markus-jakobsson.info.
`• “Using Amazon Mechanical Turk to Improve Internet Security: The Case
`of Visual Password Reset” demonstrates my understanding for novel ex-
`perimental designs to deliver and evaluate security in socio-technical con-
`texts.
`• “Practical and Provably Secure Software-Based Attestation” shows my
`ability to challenge traditional paradigms and produce disruptive tech-
`nologies.
`• “Almost Optimal Hash Sequence Traversal” demonstrates my mathemat-
`ical abilities and eye to computational efficiency.
`
`4
`
`USR Exhibit 2102, Page 4
`
`