You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Since the beginning of my interest in Artificial Intelligence, I have focused on learning and adaptation.
63
+
% because they are arguably integral to intelligence.
64
+
My central inspiration is Nature's process of evolution because it generates compelling examples of intelligent organisms and behavior.
65
+
This is why I work with evolutionary algorithms.
66
+
% and investigate their insights into generating intelligence and their application opportunities.
67
+
%An evolutionary algorithm adapts a population of solutions by means of computationally-modeled, performance-based selection, replication, and variation.
68
+
%The population supports multiple solution explorations, and algorithmic operations that combine parts of different solutions.
69
+
Neither evolution or evolutionary algorithms are efficient, but I am devoted to the open question of how to use the latter to computationally replicate intelligence.
70
+
71
+
72
+
Adversarial learning, i.e. behavioral adaptation under the pressure of adversarial competition, is an under-explored facet of intelligence, despite being so common and important to understand. For example, consider biological arms races of predators and prey, viruses and their hosts, the never-ending challenges of cybersecurity, and even squirrels versus bird feeder owners! For this, I have coined the moniker ``\textbf{Artificial Adversarial Intelligence}". Adversarial Intelligence is the thinking underlying adversarial behavior. It recruits planning, learning, technical skills, expert knowledge and other faces of intelligence. ALFA's goal became to computationally replicate Adversarial Intelligence in pursuit of Artificial Adversarial Intelligence.
73
+
74
+
Enroute to Artificial Adversarial Intelligence, we develop data-driven, gradient-based Machine Learning approaches that leverage adversarial learning. ALFA's Lipizzaner framework applies spatially distributed co-evolution to provide resilient and robust Generative Adversarial Network training. ALFA's cybersecurity projects address sponsors’ interests in anticipating possible moves and counter-moves to inform resilient defenses in cyber threat scenarios. Cyber adversaries reason with expert and common knowledge. To support artificial adversarial reasoning, I support Erik Hemberg on an ALFA open software project named BRON. BRON aggregates the text entries of open access databases which enumerate threats (their tactics and techniques and exploites), vulnerabilities (software weaknesses and exposures) and defenses. BRON supports reasoning about behavioral-level cyber threats and defenses in support of, e.g. cyber threat hunting. ALFA's results also endorse and encourage further translation of the evolutionary process to achieve intelligence different from that of adversaries, perhaps more positively, e.g. around cooperation.
75
+
76
+
Finally, veering away from evolutionary algorithms but still focusing on security, I also consider how intelligent adversaries can exploit code insecurities such as bugs or design weaknesses. Integral to this capability is how they understand code. For more on this, check out publications I have co-authored with Shashank Srikant, a now-graduated, PhD student.
Copy file name to clipboardExpand all lines: _pages/join.md
-20Lines changed: 0 additions & 20 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -27,28 +27,8 @@ permalink: /join
27
27
28
28
</div>
29
29
30
-
<!-- <div class="row">
31
-
{% for member in site.data.people.alfa_staff %}
32
-
{% if member.photo %}
33
-
<div class="col">
34
-
{% include staff_image.html photo=member.photo role="alfa_staff" %}
35
-
</div>
36
-
{% endif %}
37
-
{% endfor %}
38
-
39
-
{% for member in site.data.people.grad_students %}
40
-
{% if member.photo %}
41
-
<div class="col">
42
-
{% include staff_image.html photo=member.photo role="grad_students" %}
43
-
</div>
44
-
{% endif %}
45
-
{% endfor %}
46
-
</div> -->
47
-
48
30
ALFA respects and welcomes every applicant. We conduct ourselves as team members and representatives of ALFA with the same respect and welcoming attitude.
49
31
50
-
<br>
51
-
52
32
53
33
### MIT Students
54
34
We encourage both UROP and M.Eng students to take initiative in exploring our research areas and proposing project ideas that align with their interests and our lab's focus.
Copy file name to clipboardExpand all lines: _pages/publications.md
+6-3Lines changed: 6 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -7,8 +7,11 @@ permalink: /publications/
7
7
---
8
8
9
9
10
-
# Publications
10
+
# Recent Publications
11
11
12
-
For an up to date list of all publications (including preprints), please see [Google Scholar](https://scholar.google.com/citations?user=TgQM-B0AAAAJ) and [DBLP](https://dblp.org/pid/o/UnaMayOReilly.html).
12
+
A comprehensive list of publications, including preprints, can be found on [Google Scholar](https://scholar.google.com/citations?user=TgQM-B0AAAAJ) and [DBLP](https://dblp.org/pid/o/UnaMayOReilly.html). Keynote Addresses, research talks, and interviews can be found on [YouTube](https://www.youtube.com/@Una-MayOReilly/videos).
13
13
14
-
{% bibliography %}
14
+
{% bibliography -q @*[year>=2020]%}
15
+
16
+
<br>
17
+
See [Google Scholar](https://scholar.google.com/citations?user=TgQM-B0AAAAJ) and [DBLP](https://dblp.org/pid/o/UnaMayOReilly.html) for older publications.
Since the beginning of my interest in Artificial Intelligence, I have focused on learning and adaptation.
11
+
My central inspiration is Nature's process of evolution because it generates compelling examples of intelligent organisms and behavior.
12
+
This is why I work with evolutionary algorithms.
13
+
Neither evolution or evolutionary algorithms are efficient, but I am devoted to the open question of how to use the latter to computationally replicate intelligence.
14
+
15
+
16
+
Adversarial learning, i.e. behavioral adaptation under the pressure of adversarial competition, is an under-explored facet of intelligence, despite being so common and important to understand. For example, consider biological arms races of predators and prey, viruses and their hosts, the never-ending challenges of cybersecurity, and even squirrels versus bird feeder owners! For this, I have coined the moniker "**Artificial Adversarial Intelligence**". Adversarial Intelligence is the thinking underlying adversarial behavior. It recruits planning, learning, technical skills, expert knowledge and other faces of intelligence. ALFA's goal became to computationally replicate Adversarial Intelligence in pursuit of Artificial Adversarial Intelligence.
17
+
18
+
Enroute to Artificial Adversarial Intelligence, we develop data-driven, gradient-based Machine Learning approaches that leverage adversarial learning. ALFA's Lipizzaner framework applies spatially distributed co-evolution to provide resilient and robust Generative Adversarial Network training. ALFA's cybersecurity projects address sponsors’ interests in anticipating possible moves and counter-moves to inform resilient defenses in cyber threat scenarios. Cyber adversaries reason with expert and common knowledge. To support artificial adversarial reasoning, I support Erik Hemberg on an ALFA open software project named BRON. BRON aggregates the text entries of open access databases which enumerate threats (their tactics and techniques and exploites), vulnerabilities (software weaknesses and exposures) and defenses. BRON supports reasoning about behavioral-level cyber threats and defenses in support of, e.g. cyber threat hunting. ALFA's results also endorse and encourage further translation of the evolutionary process to achieve intelligence different from that of adversaries, perhaps more positively, e.g. around cooperation.
19
+
20
+
Finally, veering away from evolutionary algorithms but still focusing on security, I also consider how intelligent adversaries can exploit code insecurities such as bugs or design weaknesses. Integral to this capability is how they understand code. For more on this, check out publications I have co-authored with Shashank Srikant, a now-graduated, PhD student.
21
+
22
+
## Professional Service
23
+
24
+
Assuming leadership in different communities fulfills my passion for service. The scientific field of Evolutionary Computation is fairly youthful and its sub-field of Genetic Programming even younger. Since I was a PhD student, I have led different efforts to expand these fields and make them more accessible. I chaired the first workshop for graduate students at Genetic Programming 1997. When I was relatively junior, in 2005, I served as the first female chair of ACM SIGEVO's GECCO, the most prestigious conference of the field. I also was the first female co-chair of the European Conference on Genetic Programming. I was an inaugural member of the executive board of ACM SIGEVO and continue to serve as a member, having served at different times as an Officer, i.e. Secretary and Vice Chair. I am an inaugural member of the editorial board of the Springer journal Genetic Programming and Evolvable Machines, and currently am the journal's Area Editor for Data Analytics and Knowledge Discovery. The ACM Journal on Transactions on Evolutionary Learning and Optimization was founded just recently and I serve as an Area Editor for it.
25
+
I annually teach the Introduction to Genetic Programming tutorial at GECCO and recently have started to offer a tutorial on Coevolutionary Computation for Adversarial Deep Learning at GECCO. In 2013 a career recognition award propelled me to shift prevailing culture by starting an annual meeting for Women in Evolutionary Computation. To this day, it is regularly hosted at GECCO. I am especially proud to report that the meeting's scope has been updated to include any under-represented member of the EC community.
26
+
27
+
In 2013 I updated my research group’s name to ALFA from Evolutionary Design and Optimization. Based on its application focus on security, ALFA is mostly funded by DARPA and the US Government, with additional funding from multiple CSAIL Alliance Program initiatives.
28
+
29
+
Among the many rewards of leading ALFA, my group serves my passion for mentoring students and postdocs from MIT. ALFA has been a humane academic research home to numerous UROPs, Masters students (in Computer Science, Technology and Policy, and System Design & Management), PhD students, and PostDocs. For MIT's Department of Electrical Engineering and Computer Science, I have led the Machine Learning track for PhD admissions, in addition to serving on the admissions committee multiple years. I also enjoy academically advising EECS undergraduates and Ph.D. students.
30
+
31
+
Besides working for technical and academic communities, I also devote efforts to the CSAIL community. I have led the CSAIL Pretty Committee, Lobby Committee, and Logo Committee. I lead the Applied Machine Learning Community of Research and I led the committee organizing CSAIL's Celebration of its 20th and 60th anniversaries. In the service of principal and senior research scientists in CSAIL, I have worked on defining the appointments' review process. I am currently working on recommendations to CSAIL that will improve the transparency and details of the appointments' responsibilities and career paths. A detailed, full list of my service can be found in my Curriculum Vitae.
0 commit comments